Wendy Gonzalez (00:00:00):
Having diversity and having a seat at the table and representation from a huge ... I'm not saying that our thousands of people are fully represented, but it's important that we have people from Kenya, East Africa, Sub-Saharan Africa, participating in the development of AI. I think it's incredibly important that you have women helping drive and structure this data. 53% of our company is women. That was a purposeful action. Equal pay, equal job opportunity because of the building an ecosystem of the contribution to the communities. But imagine a world where the only people who sourced your data, collected your data, structured and built your models were men.

Eric Ries (00:00:43):
Welcome back to the Eric Ries Show. Today, an ethical AI company governed by a nonprofit makes the conversion to a for-profit. No, it hasn't been in the news lately. It hasn't made headlines because they did it properly. My guest today is Wendy Gonzalez, CEO of Sama. You probably don't realize this, but you've used Sama many times in your life because the people who work at Sama provide the basic training data for self-driving cars, for automated driving, for all kinds of AI systems at Microsoft, for Walmart and many others. But they're a different kind of AI company because they provide training and jobs with equitable pay and benefits to people who face employment barriers all over the world. Sama has had a fascinating trajectory as a company. As I mentioned, it began as a nonprofit and then went through an evolution to a dual company structure. A for-profit that can raise money to fund growth and a foundation with its own programs and goals, which you'll hear about later.

(00:01:41):
The foundation holds majority ownership in the for-profit, ensuring it never strays from its mission. I call this a mission-controlled company. What's interesting about an ethical company like Sama, as you'll hear Wendy describe, is that it seems like the project of a do-gooder, and yet the mission of Sama, including its commitments to its people, are ultimately what led to an extremely profitable pivot into the world of AI long before anyone saw that coming. Sama was the brainchild and lifelong passion of its founder, Leila Janah, who sadly passed away in 2020. She was a force of nature. But her vision for what she called dignified work as a way for people to not only gain their own financial independence but spread prosperity in their communities is thriving in both parts of the company. The foundation that bears her name and the company that is her legacy.

(00:02:33):
Wendy shared a lot of the details about the advantages of being a company that puts human potential and intelligence first in everything it does. And on the way I feel like she demonstrates that many of the things we've been taught about how to succeed in business aren't the only way or even the best way. The advantages of being a company that puts human potential and intelligence first in everything it does show that many of the so-called best practices of business really aren't. I know you're going to find lessons and inspiration in Sama as a model for others looking to make a similar leap. Here's my conversation with Wendy Gonzalez.

(00:03:08):
The Eric Ries Show is brought to you by Mercury, the bank account I actually use for my startup. I've been around a lot of startups and a lot of fintech products over the years. People often think the way to simplify the complexity of finance is to add layer upon layer upon layer of software and automations and workflows, and all it winds up with is a really complicated mess. Mercury's figured out the thing that really matters, the bank account. If all of your workflows and all of your automations are driven from the place where the data and the money already are, life gets a lot simpler. Mercury simplifies your financial operations with powerful banking, giving you greater control, precision and speed so you can operate at your best. We all know speed is the ultimate advantage that startups possess. Your bank account needs to speed you up, not slow you down.

(00:03:55):
Apply in minutes at mercury.com and join over 200,000 ambitious startups that trust Mercury to simplify their finances and perform at their best. Mercury is a financial technology company, not a bank. Banking services provided by Choice Financial Group and Evolve Bank & Trust members FDIC.

(00:04:18):
This episode is brought to you by DigitalOcean, the cloud love by developers and founders alike. Developing and deploying applications can be tough, but it doesn't have to be. Scaling a startup can be a painful road, but it doesn't have to be. When you have the right cloud infrastructure, you can skip the complexity and focus on what matters most. DigitalOcean offers virtual machines, managed Kubernetes, plus new solutions like GPU Compute. With a renewed focus on ensuring excellent performance for users all over the world, DigitalOcean has the essential tools developers need for today's modern applications with the predictable pricing that startups want. Join the more than 600,000 developers who trust DigitalOcean today with $200 in free credits and even more exclusive offers just for listeners at do.co/eric. Terms and conditions apply.

(00:05:09):
Thank you again for doing this. I really appreciate it.

Wendy Gonzalez (00:05:11):
Absolutely.

Eric Ries (00:05:14):
I thought maybe a good place to start would be ... And tell me if you think this analogy is strained or you find it legit. When I first started getting involved with manufacturing, I got to travel the world and see how the things we take for granted every day are actually made, and it was mind-blowing. I read plenty of books about mass production and manufacturing, and I treat the objects I get for $25 at Walmart or whatever like they're made by magic, and then they're disposable at the end. When I actually got to walk a factory floor and see just how much craft and care and human energy goes into wiring up every single little button on your microwave behind the plastic panel, it blew my mind.

(00:05:58):
Tell me if you find this metaphor legit, because I was thinking a lot about what Sama does and the fact that we are talking about generative AI and these AI models the same way. Like they're these magical things trained on data and we don't really appreciate the amount of labor that goes into the creating of them, not just by scientists who are studying model weights and transformers and things like that, but the huge amount of what they in a very sterile way call reinforcement, learning from human feedback, the data labeling and the feedback that goes on. And I wonder if you think that we've become alienated as consumers from the humans who make possible the products that we use every day.

Wendy Gonzalez (00:06:40):
Yeah. I think it's an excellent analogy, Eric. I always tell my kids it's like chicken doesn't come in packages from Safeway. It has to be [inaudible 00:06:52] and raised all these other things that happen. But AI is very much the same way. I like to think of it as the end-end supply chain for development of AI. And you made me smile when you said the care that goes into it because we see that happen. People think that AI just happens, your phone recognizes you, and that there actually is not only millions if not hundreds of millions of images being used to be able to do that level of detection so you can classify your photo, but the fact that there was likely human involved, nearly everyone in the care and attention instructions and training that they've taken to actually identify ensure things are accurate is something that I find pretty magical every day. But that's on the inside of this. I know on the outside it looks like the button you push on the microwave. So I appreciate that analogy.

Eric Ries (00:07:45):
Yeah. And I think one of the things that has always struck me about Sama is the fact that the mission is really about the humans who you touch first and foremost. And so maybe just talk a little bit about what the mission of the company is and why it spoke to you, why you wanted to get involved originally.

Wendy Gonzalez (00:08:03):
It's a great question. Well, the mission is based of a really simple and fundamental belief and that that talent is distributed equally, but opportunity is not. So Sama actually means equal in Sanskrit. And the idea was how do we level the playing field and create jobs and opportunities for people who've got the greatest barriers to employment? That's really the fundamental mission of Sama. And it really started from this notion that there's a lot of challenges in the world. Poverty being the biggest one that causes everything from wars to maternal deaths and many other just tragedies. And the first thought is how do we give aid? How do we support that? Let's build wells, let's send mosquito nets. And then not that those are not good things, but the idea behind this was that actually what's a lot better than giving aid is giving work. So creating financial independence, people can then build and invest in their own communities, and that really is the gateway out of poverty. Not aid, but work.

Eric Ries (00:09:12):
Certainly the first time I encountered Sama now many years ago, I certainly did not anticipate that one day it would be a leader in really making advanced AI possible. How did you get from one to the other? I think it's such an interesting journey and so typical of the startup journey that when you start with a real sense of mission, it's going to take you to really interesting places. How did you get here?

Wendy Gonzalez (00:09:37):
Yeah. It's really fascinating. It's this idea of purposely hiring people on the basis of impact as opposed to on their skills. Normal employment processes, where did you go to school, do you have previous experience. But we were trying to do the opposite. We were trying to bring people from the informal economy into the formal economy. So by definition that means when we're hiring on the basis of impact, meaning do you not have a job, are you underemployed, are you making less than the $2 World Bank standard for poverty, that presumes that they don't necessarily have skills coming into the job.

(00:10:10):
And so when we were looking at, okay, we want to be able to create jobs in the formal economy, the digital economy became a really obvious thing to where the challenge is there's a scarcity of jobs in region. You can do digital work from anywhere in the world. So it's a wonderful place where there's demand and then we can connect it with supply.

(00:10:29):
With that being said, we knew we weren't going to come in and say, "Hey, guess what? We'll help develop code for you." That wouldn't make a lot of sense for people who don't have experience. So it was really focused around human judgment. So how can you leverage human judgment to do what was in the olden days, everything from transcription. So take this business card and literally transcribe it into a digital format. You can leverage human judgment for that all the way to ... Do you remember Ask Jeeves from many, many, many years ago?

Eric Ries (00:11:01):
We're showing our age, but yes. Yes. You and I both remember Ask Jeeves.

Wendy Gonzalez (00:11:04):
I know what a fax machine is as well so yeah, I'm definitely showing my age. But our company's age is that in the beginning days, it was actually helping create the responses for Ask Jeeves.

Eric Ries (00:11:15):
Wow.

Wendy Gonzalez (00:11:16):
Which weren't based off of AI. They were actually based off of human responses.

Eric Ries (00:11:20):
Yeah. Ask Jeeves was the original Q&A reinforcement learning through human feedback wannabe.

Wendy Gonzalez (00:11:26):
Exactly.

Eric Ries (00:11:26):
It's so interesting how that's come back around. Ask ChatGPT is the fulfillment of that dream.

Wendy Gonzalez (00:11:33):
Exactly.

Eric Ries (00:11:33):
You're not the original founder of Sama, and I wonder if you could talk about, first of all, your background. So how did you come into entrepreneurship? I know you had had a long career in business before doing this, had worked in consulting, had worked in entrepreneurship. How did you decide to make that career change to come into what at that time I think was still a full nonprofit company, right?

Wendy Gonzalez (00:11:53):
Yeah. It was a full nonprofit. So it was quite a change. What basically it came to is that did a lot of things in my career, kept getting smaller. So I went from large consultancies, then went to some public companies, but kept getting to smaller and smaller until I decided I wanted to be a founder. So I co-founded an IOT startup, and that was great. You work so hard as a founder. There's so much you have to do, and all of the hours. I remember 14, 15, 16 hour days and then all during that timeframe, I had three kids. So three little kids under five years old. We were getting to 10 million ARR, and we were working so hard. I just remember sitting there with my husband, we would go volunteer on the weekends. We did a lot of things to try to figure out how do we show them how you become part of community, how you make the world a better place, and how to get engaged because it's everybody's responsibility? It's like you see that piece of trash on the road. You don't walk by, you pick it up. You throw it in the garbage. These are things that we need to do to show you as parents.

(00:13:02):
And we started getting really frustrated. Like wow, you have to be 13 years old to go and volunteer at the soup bank and many, many things. But at the end of the day, I got to this point to where I was like, "Okay. I'm working really, really hard to create these systems to help manage remote devices from humidors and smart televisions and smart sprinkler systems." Thinking to myself, okay, yeah, I think I'm contributing, but I'm not really, really sure if I'm contributing the way that I thought I would. And decided that it was time to actually take a look into social enterprise. And then my father passed away unexpectedly and I was like, "Okay. This is it. There's no time to lose."

Eric Ries (00:13:45):
So talk about the original origin of Sama.

Wendy Gonzalez (00:13:49):
So Sama was founded in 2008 by Leila Janah. So she had a really interesting story behind this. It all started where she gotten into Harvard, she'd gotten a scholarship and instead of going to take some of that money and her time off to go celebrate and do spring break, she decided to teach English and Ghana instead. And from that experience, she not only decided that she wanted to focus on African studies and development economics, but had basically a revelation that here were these kids that were listening to the BBC and knew more about the US political system and government that her classmates did literally. So she walked away with that. She went to school. She went to go work for the World Bank for a while. It's an amazing institution, but it is a large bureaucracy and was like, I can't really do this. It's a lot to affect change. And she started consulting and basically came up with this brainchild that there are all these outsourcing companies that are out there, but while she was consulting for one of them, she recognized that some other people that were actually supporting these large call centers, some of them came from the slums. So they may have had graduate degrees, they may have not had degrees.

(00:15:11):
And she was like, "Well, wait a second. What if we can take this concept of delivering outsourcing, but instead of hiring on the basis of your degree, instead of hiring on your experience and from top graduates, what if we dug into basically the bottom of the pyramid? The people who've got the most barriers to employment." So what are we focused only on underserved communities and focused on women? Because when women succeed, communities succeed and children succeed. If we did this, we could solve this huge problem in poverty. Because the biggest challenge is not that the people and their talent, it's the lack of access to jobs that is really the critical thing.

(00:15:52):
And so to go back to one of your questions earlier, Eric, why did I get connected to Sama? Well, that notion not only of giving work was hugely important, but I'm the child of immigrants. My husband is a zero generation Mexican immigrant who was the first person his family to go to college. And so we got to this place to where we're like, "Wow. Work and financial independence has afforded us this life." And quite frankly, the birth lottery as she called it, and we align up to it, I was looking at my kids, I'm like, "Wow. They're growing up in the Silicon Valley. They've got every chance to succeed in the world but what if they hadn't?" And so work is afforded to so many opportunities, and quite frankly, it was a lot of hard work, but it was also luck. Being in the right place, being in the right time. What if you're in a situation where it's almost impossible to even have that luck? You have no network, you have no options or access that. That really spoke to me. Beyond that, it was really about focus in Africa because in Sub-Saharan Africa ... And this is like 2023 data, almost 86% of Sub-Saharan Africa is in the informal economy.

Eric Ries (00:17:12):
Explain for those who don't know what the informal economy is.

Wendy Gonzalez (00:17:17):
So the informal economy is where you aren't salaried. It's a day-to-day job. So think of it as showing up to be a day laborer at a construction site. And in the areas that Sama works in it could be baking snacks and selling them by the side of the road, selling secondhand clothes in a stall, collecting scrap metal, and bringing that money in for recycling or bringing that in to earn money. And the challenge obviously is that number is if you don't show up to work, you don't earn any money. So you're earning less than $2 a day. If you get sick, you're not getting anything. There are no benefits. There's no continuity. So the ability to-

Eric Ries (00:18:03):
You don't even have the protection of an employment contract of any kind.

Wendy Gonzalez (00:18:03):
You have no employment contract. You have literally no protections. If you go and do the work and somebody can't pay you, you don't get paid. There's no OSHA standards. There's none of that. And so if you're unable to get into the formal economy where you have a steady salary, you have benefits, and if you get sick, you get sick days, and if you want to take a vacation, you get break days. That stability is a tremendous lift up. And that is what our thesis was. Was, hey, you do this, this helps break the poverty cycle because not only do you have sustainable employment, but jobs beget jobs. So it's one thing to get the luck to have a job, but even if you don't have that perfect pedigree of education, once you have it on your CV, I've worked in an entry level technology job. Hopefully the goal is that the next group will hire you and your ability to stay employed in the formal economy is much, much, much higher.

Eric Ries (00:19:06):
What I think is so interesting about that is we're so used to ... Those of us born into tremendous privilege thinking about low wage work as low skill work. But to make a living on $2 a day in the informal economy must require a level of determination and skill that ... I know so many college graduates that wouldn't last five seconds in that environment just who don't really work that hard. And I've never been asked to do that and yet have the sense of superiority thanks to the ... As you say, the luxury of the pure luck of their birth, the birth lottery, the tremendous education that we've all received. We're so incredibly fortunate. But on the other hand, it's cliche almost now for people to complain about the lack of grit and determination in the young people who don't want to work and to how hard it is to find people to work. And on the other hand, to praise the skill that I would guess the folks that you're talking about have in spades. So given the opportunity, I wouldn't be surprised if those skills actually, once they have the training, translate into tremendous success once they can get into the formal economy. Has that been what you found?

Wendy Gonzalez (00:20:14):
Absolutely. And that was the entirety. You've got this incredible human capital that just hasn't had a chance to shine yet. And what you said made me smile a ton because I know it is cliche. I know we've seen a lot of studies out there that grit is basically the most important skill to get hired and be successful and work. But what you said is exactly right. Our workforce has grit in spades. And they're all entrepreneurs as well. So what we've seen is when people into the program, I think there's a couple of pretty interesting stats.

(00:20:49):
So for a lot of NGOs just in history, they used to pay people to sit in training. That's a very common NGO thing. You go on, you get paid your $4 a day, you sit in a seat and you do that for two weeks. And then you may graduate, you may not. Biggest challenge is there's no job even if you had gotten trained afterwards. So that doesn't help. So that was the big model. The big shift in model with Sama was, hey, not only are we give you training, but we're actually going to hire you and employ you into the formal economy and you'll get a contract and benefits and a salary and incentives. But what was really interesting is that we took the approach of, well, we're not going to pay you to go to training. And at first this created a lot of pain. Eric, I got to tell you, people wouldn't show up. I remember right when I joined, it was going out into community centers and being like, "Hey, show up." And then we would try to get women, because our focus has always been to hire at least 50% women. And it was like we had to recruit hard because we're like, "Hey, well, we're not paying," which seemed like a really radical idea at the time. Or if we do, we're just going to be subsidy-

Eric Ries (00:22:01):
Explain the logic of it then. It seems counterintuitive.

Wendy Gonzalez (00:22:03):
It does seem counterintuitive. What we wanted is we wanted people to show up because they wanted to go to get to an opportunity. The whole goal is to say, "Hey, you get through this training, we're going to give you a job opportunity." And what's happened since then is despite the early rough days, is explaining the opposite concept, which is, Hey, we're not going to pay you to get here, we actually want you to be here and want you to try. We have a 99% graduation rate. People show up. And then once they graduate, we call them back when there's an opportunity for a job. Usually that's within weeks. Sometimes it can take a few months, depends on the cycles. And then from there, almost 85% graduate past the next training. Actually, I'm sorry, it's a little higher than that. It's actually about 89%. From there of those 89% that go into project training, which is the customer-specific things that require the next level of skill, the next 90% go. So from training all the way to getting into the production floor, the drop-off rate is less than 20%, which is pretty amazing.

Eric Ries (00:23:10):
Unheard of in these kinds of programs.

Wendy Gonzalez (00:23:13):
Yeah. So it's a long way of saying is that these jobs are super meaningful. And what was really interesting is then when we first started out, there was a lot of skepticism, but what happened was we really didn't do a lot of advertising. People just found that their cousins, their friends, their brothers, their sisters about these jobs.

Eric Ries (00:23:30):
Right. Word of mouth is strong.

Wendy Gonzalez (00:23:31):
Exactly. So it became a word of mouth. And now actually the wait list for our training program is almost 60,000. And that's not with advertisement. That's just people going, wow, you train here, you'll actually get a job.

Eric Ries (00:23:43):
Talk a little bit about the founder because she was a really special person and will be really dearly missed. Do you want to just say what it was like to meet her for the first time and her vision for the world?

Wendy Gonzalez (00:23:58):
Leila was absolutely incredible. One of the brightest lights I think that anybody could ever meet. So passionate. Like a hundred million ideas. But I very much remember my first meeting. I was like, "Okay, I'm coming from the corporate world. I'm literally thinking about quitting a startup I co-founded because this is so compelling." So I walk into the office. It's a funky office on one of the last ungentrified blocks in San Francisco. It's right in this really rough corner where there was this really rough Walgreens and a McDonald's that got shut down six months later. Anyway, it was interesting.

(00:24:37):
So I get in there and she's got this gigantic green smoothie, this huge German shepherd. I was like, "Oh." I was like, "Love your dog. What's your dog's name?" She's like, "It's Angela Merkel." I was like, "Okay. I think these are my people." Gigantic green smoothie. It was just a really, really amazing experience. And I got a chance to spend some time with her to say, "Where did this come from?" Understand really what drove the passion. And she had just this such intense, intense belief that giving work and financial dependence was going to solve poverty. And that if this idea could catch on, if this idea could catch on and we weren't the only ones to do it, we could be a pioneer in it, but we could really promote this idea of impact sourcing and hiring and underserved communities, that, boy ... If a trillion dollars is spent, that goes through procurement systems, imagine if just one of those billion was spent in impact sourcing. Imagine the amount of jobs and economies that you could create and then think about the impact of the world for Sub-Saharan Africa. It's the youngest and fastest growing population in the world. If hundreds of millions, actually billions of jobs aren't created, that could be a global financial crisis for not just Sub-Saharan Africa, but basically for the world. So anyhow-

Eric Ries (00:26:07):
Remind me of squandering of talent on just an unbelievable scale.

Wendy Gonzalez (00:26:12):
Exactly. Because I mean, it's a bit of an overused term, but after being the future, it's true. There's so much talent, there's so much activity happening, and it is the most youthful population in the world. And so to hear her passion of, Hey, you can grow this one kernel ... If we can do this at Sama ... Her pitch to me was she's just like, "Hey, let's get sustainable." So in the nonprofit world, that means earn enough revenue to cover your costs. She's like, "If we can get there, that's the holy grail milestone, and then we're going to be this proof point." And so that's really what was so incredible about her is this one idea could turn into something that was world changing and just a handful of actions could turn into a movement. So yeah, I was really inspired.

Eric Ries (00:27:05):
Talk about the decision to go to sub-Saharan Africa in the first place. Why they're among all the regions in the world.

Wendy Gonzalez (00:27:12):
I think it just started with their background and African studies. And she had met a classmate at Harvard whose father was a professor at Jomo Kenyatta University in Kenya. And so it started as a natural point of where she could get connected to talented people, youth, and have an ecosystem to work within.

Eric Ries (00:27:37):
One of the things I find most striking about the Sama story is what happened next that ultimately you made the conversion from a foundation nonprofit to a for-profit company. And that was like a niche interest area of mine. Companies that are partially or completely controlled, owned as founded out of nonprofits. That was not very interesting to most people until very recently when of course that topic blew up in a major way with OpenAI and the drama that had to do with their unique governance. I think there could be a renewed interest in the relationship of nonprofits to, for-profits. And I think one of your board members even wrote an article which we'll link to in the show notes about how to spin a for-profit out of a nonprofit the right way, unlike some issues that have come up elsewhere.

(00:28:28):
And anyway, so without getting into it, we don't need to comment on OpenAI, I know that's something we need to do. Maybe just talk about the process. You were there through that whole arc of deciding to create the Sama subsidiary and to eventually convert into a for-profit company and still yet have a role and an economic interest for the foundation. Do you want to say a few words about that?

Wendy Gonzalez (00:28:51):
Yeah. Absolutely. I think it's a couple of things. So first off, we got to sustainability and one of those pathways you'd asked about how did we get started? Well, we started in human judgment. And when I joined, we were doing lots of things. Everything from the Ask Jeeves, and ancestry.com type, we're doing transcriptions and things of that nature. And I noticed there was a couple of really interesting projects we were doing around artificial intelligence. They had to do with computer vision, training data, and at that time it was about setting up the first Xbox. Making sure that if you were to do Dance Dance Revolution, the video game would actually match your movements as well as possible. And so it was a really interesting time to where it became pretty clear to me that AI was an incredible opportunity because the need for human judgment and for humans in the loop. Now, granted back then, not everybody ... Reinforcement learning and supervised learning, those weren't common terms, but it was really clear that there was this huge need. And so we started to focus in that area. I was like, "Okay. We're a nonprofit, but we're going to take our small engineering team and our small marketing and sales team and we're going to focus it towards this area."

(00:30:16):
We grew quite a bit. We got ourselves to be, interestingly enough, a profitable nonprofit in that state. But one of the things that was really clear, especially with AI and especially with focus on data, was that we were going to need to put a lot more into the software because you just can't do this work without having a really good technical platform. So that was the conversation Leila and I had was about, hey, it's not just about achieving this financial success, it's way more about what is the future going to hold and how do we create more jobs?

(00:30:48):
We looked at each other and said, "Hey, the only way we really are able to create more jobs is going to be by taking on some fairly significant investment to really build a leading software platform and of course to grow our sales and marketing." And it's pretty interesting in the nonprofit world when you try to run after those things or like, "Hey, would you like to invest in technology?" It's not a very common thing in the nonprofit world, especially when you have a social mission or you're going to focus on the ESG components. There's always an aspect of are you aligned with that foundation's objectives that year? And so-

Eric Ries (00:31:27):
Very hard to make a long-term plan.

Wendy Gonzalez (00:31:28):
Yeah. It's not a longterm plan. It's not like a, Hey, where do you want to be? It's not patient capital. Like, "Oh, okay. I get this idea and maybe five to 10 years-"

Eric Ries (00:31:35):
It's such a shame because you'd think it could be. People always blame short-term thinking and investors on the public markets and stuff like that, yet it's the plague of the NGO sector.

Wendy Gonzalez (00:31:45):
Well, that's exactly it. I remember one year ... And this fed my conviction to make conversion along with Leila, was we had this huge foundation we had done work with, they loved our results, and they were focused on youth employment and digital up scaling great. And then the next year they changed it to-

Eric Ries (00:32:03):
You guys were right on track. Yeah.

Wendy Gonzalez (00:32:04):
Yeah. I'm like, oh, this is awesome. The next year grants up for renewal or for submitting for a new grant, and they're like, "We've changed our focus to agriculture." I was like, "Okay." So I was like, "All right. Well, we have an office that-"

Eric Ries (00:32:16):
I shouldn't laugh. It must've been really stressful.

Wendy Gonzalez (00:32:18):
No. It is funny. I look back, I can definitely laugh now, Eric, but at that time I was like, "Okay. Now how do I connect the dots to make sure that we tie back to agriculture?" So I was like, "Okay. We have an office in Northern Uganda. They're mostly farmers, but we provide digital work, which actually helps the environment because they're tilling less of the land."

Eric Ries (00:32:40):
I've written some grants in my day too.

Wendy Gonzalez (00:32:42):
Yeah. It's a classic [inaudible 00:32:42] story.

Eric Ries (00:32:42):
You do whatever it takes but yeah. It can be [inaudible 00:32:43]. Before you continue with the conversion story. I think one thing that's really important for people to understand is explain why having your own technology platform was so important and why those investments were necessary. I think a lot of people who come from a non-profit perspective would say that we'll just license the technology from someone else. Why do you need to build your own technology? That's something that for-profit companies do. But here, the impact that you're trying to drive, the software was such a necessary part of that. It wasn't something you could just grab off the shelf. So explain why that was an important thing to ... That you needed the resources to raise money to do that.

Wendy Gonzalez (00:33:18):
Oh, absolutely. So number one, we knew that to get data from our clients, because all this is about structuring and reaching data, you need to be able to ingest the data. And so having a software with flexible APIs. But beyond that, our stakeholder is the impact associate doing the annotation. So we've got to worry a ton about what is the user experience, how do we ensure quality? And there wasn't any open source or really software that was out there on the market that would allow us to do the things that we needed to do to make this a really efficient and effective process for the annotator. So that was basically our thesis. Well, we'd proven it because we'd spent some time building the software and able to manage the work, but then to do this particular work, we had to actually bring machine learning into the platform to help process the data, streamline how it was delivered.

(00:34:16):
Anyhow, long story short is we knew we needed that. Beyond that, what we need to do was to grow sales as well. How we create more jobs to close more deals. And so we're like, "All right. If we go and we actually get some investment who might also actually be able to connect us to find the right talent to engineers, to salespeople, et cetera?" So when we took a look at this, not only was there a lack of appetite to get large grants and large commitments from the nonprofit world that would sustain us and to be competitive, because after a while it started to competitive, AI became an annotation, became a huge thing. We're like, "Hey, if we don't make the investments, this could be an existential challenge. We really have to get out there and lead. We don't want to go into the oblivion."

(00:35:05):
This is important for so many reasons, not only being a proof point for the mission, but for the couple of thousand people that were already working for us. So we didn't take this as a, let's think about it, let's see how it goes. It was a deep, deep conversation. We had extensive discussions with our nonprofit board and ultimately determine that for us to be able to not just survive, but also thrive and really create the impact that we wanted to, going down the venture-funded path was going to be critical. It not only would give us access to the level of funds with, believe it or not, more patient capital behind the idea as opposed to what the foundations focuses were. And then second, it would also allow us access to the resources and tools. So all the reasons why people are looking for the right types of investors is do they have software experience? Well, that's really what we were looking for. It wasn't just about money. It was, okay, hey, we're making this transition. We need to find the right partners who get our impact, but also understand technology can help us grow.

Eric Ries (00:36:09):
So walk us through the mechanics. How does a nonprofit become a for-profit company?

Wendy Gonzalez (00:36:14):
So it is a very interesting process. So there is a formal process. And because we're incorporated in the state of California, we need to work with the attorney general's office. We had to submit a specific application and basically prove that as we are making a transition from non-profit to for-profit, that there would not be any excess benefit to the nonprofit. Because for years, it had been getting basically supported tax-free from the state of California, and actually a nonprofit is owned by the people of the state of California. At the end of the day, the ownership is really to the people, not to the individual. So going through a non-profit conversion, what was very interesting is that while we went through all the checks and balances ... It was a really fascinating process, Eric, because most companies who do this are bakeries and catering companies. Very small businesses, rarely above a million dollars. We're like, we're a $15 million technology company and we want to make this transition.

(00:37:16):
So it took six or seven months and huge amounts of documentation. And at the end of the day, we had to separate all the assets between the nonprofit and the for-profit. And the key was effectively that we made a decision in creating this moving from non-profit to for-profit, that effectively we separated the businesses assets from the non-profit programs. And then in exchange for basically setting up this for-profit, the nonprofit got a hundred percent ownership of those assets. So it was almost like creating some subsidiary.

Eric Ries (00:37:56):
Yeah. Yeah. I think it's important for people to understand when we talk about conversion, there's actually now two entities where before there was one. It was much more mitosis than it is like a replacement. And the foundation as the owner of the subsidiary has the stewardship responsibility to make sure that it stays on mission.

Wendy Gonzalez (00:38:14):
That's exactly right. So it is not exactly shutting down one company and then starting another. And it's not like a one-to-one conversion. They call it a conversion, but it really was split into two companies. And then the for-profit company, Sama basically became almost like a subsidiary of The Leila Janah Foundation. And what it really ultimately meant was that the foundation was the a hundred percent shareholder of Sama. And in that process of making the conversion, just to get a little bit more technical here, there was a requirement to not have any excess benefit. So what I mean by that is that typically when a founder starts a business, if it's a for-profit, they have a hundred percent of the shares. In this case, because we started as a nonprofit, the state allowed us to designate a small portion of those shares that were created to the employees. But it was a very small portion. It was sub 15%.

Eric Ries (00:39:15):
And then the for-profit investors said ... You went to Venture Investors. They invested only in the for-profit subsidiary, and that effectively diluted the foundation's ownership down from 100%.

Wendy Gonzalez (00:39:27):
That's exactly right. So through series B, Leila Janah Foundation, the foundation is still by far the large shareholder, and it probably would be through series C. And for a while from there.

Eric Ries (00:39:39):
And the foundation appoints members of the board. How is the foundation's interest represented in the governance of the company?

Wendy Gonzalez (00:39:45):
So, yeah. So when we set that up, this is one of the reasons why I also thought it was clever, a belt and suspenders, is that with a large shareholder, you get representation on the board. So we have a member of the foundation that sits on the board, which is key. A voting member who gets to vote on, of course, all of the important items. The other thing that we made clear to our investors as well is that as soon as we were going to make that conversion, we would become a certified benefit corporation. So that was actually part of their understanding coming into this that we would also want to put those clear independent requirements into this. And beyond that, we decided, okay, well, on top of that, why don't we also find some impact investors? So we have a combination of, for-profit investors, impact investors.

(00:40:36):
And so our board makeup basically is an independent board chair, a couple of for-profit VCs from series, A, series B, and our foundation chair and myself. So it strikes a really nice balance. And also because the foundation's a shareholder, you don't get a lot of misalignment. So we all want to do the right thing, which is to honor this mission and grow the value of the business. And that was the brainchild that Leila had, and we spent a lot of time talking about before her passing was, if we do really, really well here, if we can create this really valuable business, prove that you can grow a successful business, be very clear and stick to your impact, values and mission and hold our operating model, not only are we this huge proof point, but then the foundation gets the benefit. At the end of the day, this becomes hugely valuable, the foundation then gets the benefit, and the foundation's mission is to spawn other businesses like Sama, to continue to create enterprises that give work.

(00:41:47):
So I've got a list of literally about a hundred next Sama businesses that we could start, should we get a really successful exit. So that was the idea is that we're all marching in alignment. And so I think when we talk about where's there a right way to do it sits in the governance. And beyond the governance, do you have a shared vision or shared outcomes? Yeah.

Eric Ries (00:42:15):
That to me is one of the most important and neglected aspects of governance. There's basically the two dimensions of governance that I think are so important that people don't give nearly enough attention to are coherence and integrity. And Sama to me is just such a beautiful example of both. You have to have that feeling of alignment that the whole organization is pulling towards some purpose but without which it can't achieve extraordinary things. It's not possible. But then integrity is also really important. I think about you're building a product that spans continents, that requires cooperation between people who may never meet and would never have met, if not for Sama. Were bringing so many stakeholders together from so many different places in an industry, especially in the outsourcing that has a history of abuse and exploitation and a lot of bad behaviors, and that has cut both ways.

(00:43:05):
It has meant that people who are employed in those companies have had bad outcomes, but also plenty of companies have felt defrauded, cheated. They can't supervise the work product, they don't know what's going on. So trustworthiness is such a powerful asset to have on your balance sheet for all those people to trust that you will help them achieve their goals if they partner with you. It requires making long-term commitments. It requires people to reciprocate that long-term commitment, and it wouldn't work, I don't think, if it wasn't such a high integrity team, if investors could show up at any time and just knock you off mission or force you to betray the promises that you've made, nobody would believe those promises. You wouldn't last five seconds. And yet, by following this path, it's led you to a lot of business success.

Wendy Gonzalez (00:43:53):
Yeah, that's a hundred percent it. Because if you don't have integrity and you in particular, what I'd say is don't have investors that align towards that, I think you're cooked from the beginning. So not only identifying the right investors, but being as upfront as possible that this is our committed social mission, and we let everybody know, by the way, beyond being a B corp, we are going to transition to be a public benefit corporation. So then even as part of that legal framework, there's an opportunity to raise items as it relates to the impact. So there's a mechanism if there is really any concern over any conflict between impact and profit.

Eric Ries (00:44:38):
This episode is brought to you by my longtime friends at Neo4j, The graph database and analytics leader. Graph databases are based on the idea that relationships are everywhere. They help us understand the world around us. It's how our brains work, how the world works, and how data should work as a digital expression of our world. Graph is different. Relationships are the heart of graph databases. They find hidden patterns across billions of data connections deeply, easily, and quickly. It's why Neo4j is used by more than 75 of the Fortune 100 to solve problems such as curing cancer, going to Mars, investigative journalism, and hundreds of other use cases. Knowledge graphs have emerged as a central part of the generative AI stack. Gartner calls them the missing link between data and AI. Adding a knowledge graph to your AI applications leads to more accurate and complete results, accelerated development and explainable AI decisions all on the foundation of an enterprise-strength cloud database. Go to Neo4j.com/eric to learn more.

(00:45:45):
We are all taught and indoctrinated, I think is the right word, that business, capitalism, competition seeking profit is an inevitable race to the bottom. That at the end of the day, you're going to treat people as badly as you can get away with. And here we're talking about, so-called lowest common denominator labor that should be a complete commodity and can be easily replaced. We think about the history of globalization, of going to one region after another and seeking the lowest wages and the least worker protections. Therefore, according to that theory, your commitment to treat people in this respectful way should be a major liability. In fact, the success of Sama is impossible according to these theories because you would face competition from people who would undercut you by price and would drive you out of business and yet here you are. So I'm curious, you're living proof that there's something wrong with the theory that we've all been taught. And I'm curious what your experience has been like having this extra requirement that you have to operate against, and whether you feel like that's been a competitive advantage or a disadvantage.

Wendy Gonzalez (00:46:59):
It's really interesting. I've felt at times both, if I being honest, Eric. I've felt at times both because there's no question that there's pricing pressure and there are going to be those clients and those companies who want to push a race to the bottom. But the way that we really tried to approach this ... And I think it helped being a non-profit first, honestly, is that we had total clarity on our mission. We're committed to living wages, we're committed to formal employment and contracts. Like, everybody gets vacation benefits. These are a critical part of what we're doing because our whole goal is to create a place where people can begin to build their next careers. And because we had such clarity, it became really obvious to me that we cannot be ... There's the, are you going to be the cost leader? Are you going to be a premium company? Well, it became really obvious we need to be on the premium side. We need to be in a place to where we can actually command a higher price point, and how do we build that both based off of reputation and brand, but also this amazing quality product.

(00:48:06):
So we came into this not looking at, okay, wow, we've got to compete for the cheapest, but how do we elevate the value of the human envelope? What really is our real differentiation in this model? And I would have this conversation often with clients. If I can delve just a bit deeper into the AI aspect of things, it'll hopefully bring this to a point, which is okay, a lot of AI is built through the crowd. It's a revolving door of people. You go on Amazon Turk or whatever app with some of our competitors applications, and you're going in, you're getting a set of instructions and you're getting a task. And those tasks can be paid cents on the hour. They get paid more. Depends on how much demand there is. And I was like, "Okay, that's great. But ultimately, at the end of the day, if you're trying to build AI that is extremely high quality ..." And for sensitive applications, autonomous driving is a great example. That's a safety-based application. You can't be accurate 92% of the time, four out of five times you got to be right almost all of the time.

Eric Ries (00:49:13):
That definitely is not.

Wendy Gonzalez (00:49:15):
So this idea of being able to say, "Hey, we've got to workforce that's not dedicated to you, but they're trained and they're trained on your specific application. And as your model evolves, we stay with you. They retain that knowledge and we pay people well so these are super meaningful jobs. So our tenure is three and a half, four plus years. Most of the people who come in and we've got much larger tenure, it's because they move up. So here you've got this valuable set of expert resources that are dedicated to you, and that value is so much greater than something you get transactionally. And yes, that transaction could be a lot cheaper, but you're not going to get the value out of it."

(00:49:54):
So that's an example where we had to say, "Hey, how does our business model really create a more valuable product for our customers?" Because you can take that exact flip thing. So this is what we communicate and we say, "Hey, this is why we actually command a higher price point, but you're getting way more business value out of it." So I've been the first to say, we're not for everybody. If you want something where you need a million people for one day, we are not going to likely be the right solution for you. But you need to build state-of-the-art in ML or state-of-the-art, gen AI we've got this incredible workforce, this incredible platform.

(00:50:38):
And we could have looked at it the other way right Eric. We could have looked at it as, okay, this is a massive liability. We've a commitment to a living wage. And over the past couple of years, inflation has been horrendous. In particular, we operate in East Africa. Two years ago it was almost 18%. Last year it was almost 14%. And you can imagine that while we have our commitment to living wages and we meet the those and we do location-specific evaluations, we honor those living wage adjustments. And it's not just the entry level because if you raise the entry level, levels above that also have to be raised because you don't want to create a lot of concern about compression and things like that. So in that situation, are clients going to turn around and give us 14 to 18% increases? Well, probably not. So what do we do? We have technology. We drive value that allows us to still be successful as a business because we are able to not just go on an hourly basis and erase the bottom, but have different commercial mechanisms to that, yes, we have to make investments in our business, but it's also true to exactly what we communicate, which is we care about having a continuous workforce that is growing and dedicated to you.

(00:51:59):
So it's a long way of saying you can turn what seems like a liability into ... It's really not a liability. It's really a business strategy. And we then explain what is the value of that business strategy? Like I said, it doesn't mean it's for everybody, but I think if you don't look at it like that, if you don't align what your social mission is with how you drive your strategy, it's not possible to operationalize it. Because if we were stuck in the other view, we'd say, "Sorry, guys. Clients only agreed to a cost of living adjustment to about 2%. So even though I know things raised by 18%, we're just going to stick with two because that's what we need to keep everything even." And we don't do that. We strive towards ways to get to the premium. We recognize that some of the investments we're going to have to make are going to be longer term. They're not going to be in quarter or in year. And that I think has ultimately created as you actually reference the reputation and the integrity of, okay, well yeah, we can actually believe you because you're walking the walk.

Eric Ries (00:53:08):
Mm-hmm. What's so interesting is I am thinking about generations of business strategy consultants. Going back to Deming who it used to drive him crazy. Peter Drucker used to complain about this all the time. It's a classic. Tom Peters complained about it. Jeff Moore. I can think of so many who just have been driven crazy by the fact that businesses race down to build commodity products, and they don't try to improve margins through value. They just think about cutting costs and therefore they erode their own competitive position. Deming used to think this was insanity, that as you lower the quality of the inputs, you lower the quality of your product, though probably you lose your competitiveness and it gets you into this downward cycle.

(00:53:46):
So to me, it's really fascinating from the outside to see the ways in which your mission has forced you to pursue higher value over time, increasingly high value areas for your application, which has led you into the hottest area of them all. Whereas other people would be at pains, I think, to say, "Oh yes, we saw the AI revolution coming, and of course anticipated the new AI and how it was going to work and all the things that it's needed." You didn't need prescience for that. You had the mission. Just like I said forces. It's funny, such a funny thing to use the word forcing to describe, but you were forced to seek out these higher value applications, higher value opportunities, and that's now a source of tremendous competitive strength.

Wendy Gonzalez (00:54:30):
Yeah. It's interesting. And that's exactly right. It was a little bit of a forcing function because that, that's where you got to be. If you want to have both. You have to be in that upper right quadrant. You can't be on the other end. And it was really fascinating because it shaped our technology. So when we started, when we built our technology all those years ago, we were like, "Okay. We need to worry about dealing with enterprise and mass volumes of data and having this super scalable system. Yet at the same time, we've got to worry about the human in the loop. The person's actually using it, our annotators, our impact associates." So we ended up putting in lots of effort into quality assurance, lots of effort into usability and all of the alchemy, we got to be modular and flexible and recognize that the data is going to change all the time, all of those decisions, because we're also looking at our annotators, like the stakeholder brought us to our differentiation today, which is we have the best quality, amazing analytics.

(00:55:33):
We've leveraged the technology to do the heavy lifting that allows really our humans to maximize the value of their judgment, basically. And so you tie that into generative AI and here you've got different things that you're looking for. It's more model evaluation, of course, more supervised, fine-tuning, but you've got to control for hallucinations. Machines can't yet check machines to see if they actually are building the content that is intended. You need humans to check that. And so that's just another example of how this notion of human judgment AI and the value of processing mass amounts of data and having a platform to do that too, which puts us in a really unique position to where we don't have to say, "Hey, our goal was always to automate it all away." It's like, well, no. We've always had this belief that human judgment is a critical part of AI development. We're embracing that.

Eric Ries (00:56:29):
I think everyone's heard the phrase human in a loop lately, and if not, believe me, you will. The mission took you to that ahead of a lot of other folks. Generative AI I think is really interesting. We've already had several scandals and debacles caused by generative models going off the rails in various ways, and this is before we even get to the real safety debates, as the models get more capable and get wired up to more sensitive systems, how dangerous. But even trivial things like the fact that the certain models overuse the word delve and certain other keywords that are such a tell that ... At least from my perspective, it just seems like these companies are shooting themselves in the foot. You could just tell the problem is they use low quality feedback. If they had paid the premium to get high quality people in there to actually give the thing better feedback, it wouldn't have made these mistakes.

(00:57:18):
By treating the model as the source of the intelligence, we miss the fact that the quality of the human intelligence that touches it is really where the value comes from. And these companies are spending hundreds of millions of dollars, if not billions of dollars training models, and yet are outsourcing the really critical human feedback part of it to humans who are working in terrible conditions, who are not properly paid, who are under tremendous stress, who do not have the training and the education to do it properly, and they're reaping the benefits. I should say reaping the whirlwind. This is the thing that Deming was railing about all those years ago. You squeeze your suppliers and you get what you pay for. And I just think it's really, really interesting to see how catastrophic those errors are turning out to be, even now just the dawn of this new era. But I think as the stakes get higher, there's going to be much more of an emphasis from the public, from consumers, from the enterprise customers of these models to know, was this data ethically sourced? You're talking about a data supply chain, but that's like what we're talking about here. Are we going to need standards and ethical guidelines for how this should be produced in the future?

Wendy Gonzalez (00:58:29):
Yeah. 100% believer in ethically sourced or the ethical supply chain. It's like you've got to know at the end of the day, all this is built on data. You need to know where did the data come from and was that okay for you to use that data? So you have the data provenance component of it. Beyond the data provenance component, then transparency and well, who structured your data? Where all the value of the data comes from, is who structured or labeled your data or who evaluated the model that was generated by synthetic data?

Eric Ries (00:59:05):
Yeah. Who did [inaudible 00:59:06] teaming? I'd like to know.

Wendy Gonzalez (00:59:07):
Yeah, well, exactly. And were they paid where they paid fairly? Think about it, it's like the cacao farming. It's like eventually, do people feel like it's okay to eat a chocolate product that was ... by a child? And all that came to light? Well, I think the same things are going to need to come to light and not just because of the ethical reasons, but also because of the outlets. So imagine this ... And this is a common thing with the crowd. If you are driven on a task basis, imagine you're working on a platform and many of these platforms, they do surge pricing much like Uber or much some of these other applications where imagine you've got a task of all these people competing for it. You start the day, the task is worth $4, you end the day and it's worth like 20 cents because now all of a sudden people are running after it. What is the objective when you're doing something task based? Well, it's to complete the task. Because you don't get paid if you don't complete the task. And so therein lies the exact challenge.

(01:00:12):
I was talking to a friend and she was like, "Oh, yeah ..." I was explaining to her what I do and she's like, "Oh, yeah, my neighbor does annotation." I was like, "Oh." She's like, "Yeah, she just does it for fun." I was like, "Oh, well that's really interesting." "Yeah, she tells me she goes there and she doesn't ... She keeps clicking on it until she gets it right and then keeps clicking to what she thinks is going to be right." And I was like, that's exactly what all the studies say is that when you're going and you're getting paid only if that output is successful, you can end up with a crowd bias, which is to say whatever answer you think is going to be right, so you can actually get paid. Imagine that versus an employment model where your incentive is, yes, of course you've got to worry about productivity, but your success is on quality. You get paid whether your task completes or not, but oh, by the way, you get incentives for quality and meeting these other objectives. So it's a very, very different model. And I think when you think about, okay, we're building technology that is ubiquitous, right? It knows no boundaries.

Eric Ries (01:01:15):
Absolutely.

Wendy Gonzalez (01:01:17):
Do you want a safety application? Do you want-

Eric Ries (01:01:22):
I'm just thinking about self-driving cars self-driving cars. You get in your self-driving car, and do you really want it to have been trained on someone's data who was just clicking to get to the next screen so that they could get paid? I don't think people as consumers are yet really aware of how important quality is as an input to this production.

Wendy Gonzalez (01:01:39):
That's right. And exactly. It's the same way with traditional ML. It's the exact same way with generative AI is that if you don't have those checkpoints, that tuning those quality checkpoints, yeah, the outputs that you get may not be what you expect. And so having the human, a human is going to be able to not only validate for accuracy, but taking a look and understanding the data is going to allow us to understand bias, and that's huge. So having that visibility all the way through to say, "Hey, this product was made and it's safe to use, if you will." You can't just have a label that says generated by AI with nothing else. Eventually you get to the hairdryer label, says like, don't plug this in next to water. Well, you need that level of visibility that says, this is what it works for, this is what it doesn't. And if you want to see, there's visibility into how this was actually built. I think as a consumer, people affected government should be worried about that. The EU Policy Act, I think did a good ... They made some good steps towards identifying potential sensitive areas, but they did reference both responsible and ethical AI and hopefully we'll continue to see more of that, the executive order from the Biden administration.

Eric Ries (01:03:05):
Yeah. They have a few things to work on still there.

Wendy Gonzalez (01:03:12):
Yeah. No. I was wondering if it was going to get closer to where the responsibility act one, but yeah, it's got some ways to go.

Eric Ries (01:03:21):
Yeah. Well, and everything's so nascent and the level of uncertainty is so high it's a really tough area to regulate. I'm going to go back though because just know there's some people listening to this, I just know it, you're out there. I can just speak to that person just for a second. I know you're there who's listening is just like, okay, do-gooder, woke, whatever, ESG. That's become such a flash point. So polarizing. And we talking about the ethics and the responsibility here, but you said something that was really important, which was that apart from the ethical reasons to do it, there's also a strategic reason. It's a key part of the company's strategy that has been a competitive advantage for you over the years. So talk about the for-profit case to do this.

Wendy Gonzalez (01:04:10):
Absolutely. So much of the time spent on building models is on structuring and evaluating data. You want to get the most out of your data. You want to amplify that. If you better understand and have more improved quality data, you actually accelerate your development development cycle. So we want to help our clients basically bring their models to production faster, and you can spin around and around ... I'll use a couple of very simple examples and maybe some more complex ones, but if you're building a self-driving car and you've got all this unstructured data, maybe the car drives really well in the US, but it doesn't do so well in rural areas. Maybe the ML model does extremely well in Germany, but could not recognize 14 lane Chinese roundabout. So the data's going to be constantly growing, changing. You need humans in the loop, even as your model evolves to help identify what those edge cases are because by definition, your model doesn't. So you've got humans loop who do the edge case detection.

(01:05:20):
If you have a set of employees or you have what I would more I'd say describe as a closed feedback loop, so humans loop who's seen your model and worked with your model as it evolves, their ability to understand your business rules and identify those edge cases is very high. The quicker you get that feedback loop, the quicker your model is going to ... You're going to annotate new edge cases. Your model's going to get smarter and smarter. By having also humans evaluate the content of what's sitting inside your data we can ingest all of this information into our platform and see that, wow, your model's not doing well on nighttime images. You don't have nearly enough nighttime images. Let's talk about you getting more of the data. So those feedback cycles, because it's an iterative loop, save a tremendous amount of time, and at the end of the day, the goal is to not maximize the amount of data that you're training. It's actually to minimize it. So if you do it intelligently and thoughtfully with a smart group of humans, loop in a platform, then you'll get there much faster.

(01:06:23):
So we've actually looked at it as how we can save our clients money and get their products out the door faster. It's no different with gen AI. Whether it is validating to make sure that the model's working as intended and to avoid hallucinations and bias or to gen AI models that are in production. The gen it gen AI, because it's constantly learning and growing, creating new content. So even if you deploy the model, it still needs to be checked to see if it works against the originally intended parameters. So here you've got people who really can understand your model and that time to be able to react and identify gaps is much faster. Well, aside from the ethics side of things, just talk to the ML engineers who are building the applications. They'll tell you that that's exactly the thing that you need to focus on.

Eric Ries (01:07:20):
So your mission has given you a sneak peek into the future. You were on AI and machine learning and Gen AI like long before the rest of us. So what do you see as coming in the future now? What are your expectations based on what you're seeing? What do you see as the next wave of AI enhancements? What should we look forward to?

Wendy Gonzalez (01:07:45):
Besides Skynet?

Eric Ries (01:07:45):
Yeah. I hope besides that, yeah. Are you worried about it?

Wendy Gonzalez (01:07:50):
Not for the next few decades.

Eric Ries (01:07:53):
As the one who has to annotate the data to find out what these things know and don't know? you're not-

Wendy Gonzalez (01:07:57):
Exactly. Exactly. But I think the thing that we see coming forward if we're really excited about is the converted convergence of multimodal models. So we see a ChatGPT, we see video creation models. Well, it's all going to begin to converge to where you have text to image and video, and it's going to completely, I think, change the way we interface and create AI. So we're already seeing some really interesting applications in practice. That's going to be huge. It'll affect every industry so well beyond copilots. You're now going to have the ability to do that with images, video, LiDAR data. The next is going to be atmosphere data. You name it.

Eric Ries (01:08:51):
Yeah. People are sleeping on multimodal models and what they're capable of, especially because people can a little bit imagine, okay, ChatGPT but with audio and text and video. But we're starting to get into more of a degree. It's not a matter of degree, it's a difference in kind. When we're talking about LiDAR data, 3D data app, atmospheric data we're talking about models that can really reason about the real world and can guide scientific experiments that can operate machinery and do automation in the real world that can interact with people to amplify the work they're doing, not just the knowledge work sitting at a computer, but even in factories, in construction, in basically every field of human endeavor. It's about to get real interesting.

Wendy Gonzalez (01:09:36):
Yeah, I think what I get really excited about ... I did a talk with folks at NASA a while back. There's some pretty incredible things that can be done when you start to think about the convergence of all those forms of data. And then there is ... Actually, manufacturing is exactly where my head goes. I think there's a huge amount to be done there. But even the really basic examples to where we're building autonomous level one, two, three, four capabilities for self-driving cars, but imagine really where the great adoption is going to be in level two and three where there are humans in the car. Well, it could be as simple as giving voice commands to your car to drive. There are just so many very practical applications and then literally otherworldly applications.

Eric Ries (01:10:29):
Okay. Just a quick lightning round here at the end. You very famously talked about what it was like for your employees to have to do Facebook moderation and to do that work before you really made the pivot into AI. Just talk about the decision to do that and then to stop doing it.

Wendy Gonzalez (01:10:46):
So it was definitely an exception project for a client who was like, "Hey, we want local people to do moderation for these languages because you can either outsource it to different part of the world, like the Philippines or otherwise, or have local people doing local content, and that's much preferred." We're like, "Hey, this isn't our business model. Our focus here is on annotation, but it makes sense." The teams felt a sense of pride to be able to support their communities. But it was a very, very isolated thing. It was not our impact model, because to be able to do that, you need language skills, you need resilience. Only certain people could really support this type of work. Was completely different than what our core annotation work was. So it was certainly quite an experience doing that. We ultimately came to a conclusion that one, it's sucking all the air out the room. A couple of hundred people against our 4,000 plus workforce. It's not impact. It's not the core of what we do, and it's also quite frankly, distracting from basically our core model and also from our social mission.

(01:12:14):
And so while it was a separate business, if you will, there were a couple of things that came out of it. One was it just didn't make any sense to not focus on core ML and annotation. That's what we do. That's what our platform is. And it was not effective for the business. But beyond that, it was distracting. And one of the things that we did that was important as a learning from that was we basically said, "Okay. That was an exception. It's not something that we pursued. How do we make sure that we don't end up in a situation again where we could end up away from our focus?"

(01:12:58):
So we basically went to the board and said, "Hey, we're going to create something called a service line boundary of work that we will ..." I thought it was always unspoken about what we would do and not do, but we need to actually speak in and write it down and get it up into the board level. So we created boundaries of the kind of work that we will not ever do and accept, and that is everything from big tobacco and weapons, to use sensitive content, pornography, you name it. So we wrote it all down on the list, and then we said beyond that, what we're going to do is we're going to create an ethics guild. So we create an ethics guild that includes our impact associates, so representation from across the company that doesn't have any management in it. That basically is the group that says, "Hey, any opportunity that comes in, whether it is inbound or otherwise, can be evaluated by the ethics guild to see if it draws anywhere into the gray area." Because we already clearly know what we're not going to do. But things change. What happens now? Many, many years ago, AI wasn't necessarily being used for surveillance. So then we're like, "Okay. Well, today we say we're not going to do anything that is surveillance related." Well, what ends up happening if that changes a shade where the data is being ... Issues change over time, basically.

Eric Ries (01:14:20):
Sure. A fine line distinction between self-driving and surveillance.

Wendy Gonzalez (01:14:24):
Well, exactly. So we're like, we need to have the flexibility to evolve this as it goes. So we're going to bring together a team that can evaluate anything that goes around the gray and give ourselves the freedom to be able to add to that. And so it's something that we actually write into our contracts. We're like, this is just the work that we won't do. So one big lesson learned was that we have to be completely clear, and so we brought it up to the board. We've implemented that, and now that anybody in the company can say, "Hey, I'm not sure about this one. This feels like it's falling into the gray." Or I should say, not we because I'm not involved in the guild, but they gave me a heads-up that, Hey, we're going to review this case. So anyhow, I think that at the end of the day, that is where we were, but I think at the end of the day, we actually, it helped drive some clarity. We were able to put some, I think, better practices in place as well.

Eric Ries (01:15:18):
As an aside, have you thought about encoding the guild into your governance, either as a committee of the board or as an employee voting trust or something like that?

Wendy Gonzalez (01:15:26):
Oh, that is a really good question. The boundaries are in our board governance, so it had to be signed off by all of the boards so everybody is aware of what the process is, but is the guild itself as a mechanism? I think that's a good one to follow up on.

Eric Ries (01:15:47):
We can take that offline if you ever want to do that. I have some experience with those kinds of structures.

Wendy Gonzalez (01:15:51):
Yeah. Yeah.

Eric Ries (01:15:52):
So it couldn't be undermined in the future if you change your mind.

Wendy Gonzalez (01:15:55):
Yeah. Yeah. No, that's really interesting. I thought we'd had it covered by having it into the approved through the board policies and practices, but yeah. Yeah. I'd love to follow up on that.

Eric Ries (01:16:09):
Whatever. I don't mean to put you on the spot about, it just occurred to me as you were talking about it. What a great idea.

Wendy Gonzalez (01:16:13):
Yeah. No. It's a good one. I think that's part of it. It was really interesting. Just as a side note, going through the board and saying, "Hey, we feel strongly." I'm like, "Hey, I don't want to be in a situation where something, especially as we grow and scale, that something gets through the cracks or we haven't made it completely clear because this was a decision. We're moving away from it. We need to be totally clear." It was interesting to bring that up to the board to say, "Hey, we've got a group of people and we've got a policy that says we can decline. A 10 million deal could come in, but if it came in under this, we wouldn't take it." It was really interesting because this is where we're getting alignment with the board becomes so important is when we stated these and said, "Hey, we need to stay very true to our mission and goals." And even in doing that moderation work, we weren't hiring impact people.

(01:17:07):
So, hey, we have to divest this work. We need to focus over here. It was a really interesting conversation. If you don't have people who get the social mission part of it could have been brutal, but instead, well, it took quite a bit of conversation, "Well, what does this mean? So, wow, you really do have people that are outside of you that can basically say no to work." We're like, "Yeah, but this is why we believe it's important." And ultimately, they lined up and signed it into practice.

Eric Ries (01:17:36):
This such a great example. Founders are always telling me they want the one weird trick to have their governance be set up permanently and forever with no problems. And it's like even here, you've gone to such lengths over so many years to get it right. Still, it's a continuous practice of discovering new ways, new cracks in the foundation, new things that might get in, and of course, encountering new situations in the world that require nuance and judgment really to get it right so I applaud you for doing that.

Wendy Gonzalez (01:18:03):
I appreciate that. And also honestly, I think our employees demanded that too. They're like, "Hey, we've got to hold ourselves to a higher standard." And so it was a really great thing that was not just a handful of people. It was a collective, this is the right thing to do.

Eric Ries (01:18:21):
They're the most invested in the success of the company, more even than your investors, I would say.

Wendy Gonzalez (01:18:25):
Yeah. Yeah. At the end of the day, I think it's made us definitely stronger for it with better-

Eric Ries (01:18:32):
I'm very conscious of the fact that this whole conversation, I've been asking you questions really from an American perspective, it's natural. Here I am in California living my very nice life. I get to have conversations with people like you. It's extremely nice and very interesting. I'm curious if you have anything you'd want to say to your employees and their communities back home. If any of them were choose to watch or to listen to this, what would you want them to know?

Wendy Gonzalez (01:18:54):
Oh, I hope some people do. Actually. I just came from a couple of weeks in Kenya and Uganda so you caught me. Hopefully I'm not overly jet-lagged, but I literally just came back. Yeah. What I would say beyond what we've shared before is that this is really all about empowerment and building skills for the future. So one of the things that really excites me about the work that we do is that it does change. What we did 10 years ago we don't even remotely do. What we did even three years ago in AI or five years ago. Is there a dog or cat in this picture? And now people are dealing with 3D complex LIDAR, 2D, evaluating these models that and generating really complex responses to some of these generative AI models. What I get excited about is that for all the change, which can be challenging, what we're doing is we're building skills for the future. And every time we succeed, we are in a position for somebody to have a positive outcome, which could be leaving the company, which I know sounds like the opposite of what you want. But if you can create those skills and somebody can leave, that's a positive attrition. So I think it's coming upon us to continue to invest, and while we challenge what we've seen from our teams is incredible.

(01:20:15):
Eric, I promise you, if you saw some of these annotation tasks or model valuation tasks, I have trouble completing them. I couldn't get my task passed. These are incredibly talented people. And so the biggest thing is really about investment in growth. And probably the last thing I would say, I guess is that when we talked about the supply chain, I think this is not something talked about a lot, but a lot of artificial intelligence is ... Well, it's put right here in California by a handful of people and a certain demographic. A lot of people who might look and be the same.

(01:20:50):
And when you think about this technology, well, this technology isn't just going to be used in California. It can be used all across the world. So having diversity and having a seat at the table and representation from a huge ... I'm not saying that our thousands of people are fully represented, but it's important that we have people from Kenya, East Africa, Sub-Saharan Africa, participating in the development of AI. I think it's incredibly important that you have women helping drive and structure this data data. We have 53% of our company is women. That was a purposeful action. Equal pay, equal job opportunity because of the building, an ecosystem of the contribution to the communities. But imagine a world where the only people who sourced your data, collected your data is structured and built your models were men or men from California or whatever.

Eric Ries (01:21:43):
I don't have to imagine it because we have a serious problem in the industry. I live it and I know you do too. And apart from the moral and the ethical imperative to do something about it's also an economic imperative. We literally don't have enough talent working on these important problems. We don't have enough startups being built. We don't have enough technology. As wondrous as these technologies are, they're being built by a very small number of people, and the demand is way outstripping the supply. And when that happens, you just get price inflation. You don't actually get the creation of more value. So I view it absolutely as an economic strategic moral imperative to diversify who is doing it. So I agree with you there completely.

Wendy Gonzalez (01:22:23):
Yeah. And if we get to be just a small part of building an ecosystem, it'd be incredible. So they've got a startup community in Nairobi. This place called The Garage. We've had folks who have successfully left our company to go be software engineers. We run a business grant program for impact entrepreneurs. So it's like basically it's where we get to be investors for a couple of days a year and fund impact entrepreneurs and-

Eric Ries (01:22:56):
The most positive of positive attrition. That's great.

Wendy Gonzalez (01:23:01):
Yeah. And some have built these awesome apps and have left, but some have created companies that make shoes out of recycled rubber. All sorts of things. But the entirety is like, hey, you build that ecosystem and that energy, that mojo is incredible. And it shouldn't be concentrated in one part of the world. And then the second piece of it is why shouldn't everybody in the world be able to get benefit from this AI economy? So it's like there's so many reasons why I think it's important, and that's basically what I would share with my teams is that you deserve your seat at the table. Our job is just to make sure that you can get to the table. That's it.

Eric Ries (01:23:44):
All right. Final question. If someone right now is listening to this and they're thinking about starting a company, they want to create change in the world, they have a long-term ambition, a vision, what advice would you give them? What do you know now that you issued down 10 years ago, and how would you advise them to encode their change in the context of a for-profit company?

Wendy Gonzalez (01:24:06):
Oh my gosh. Well, the first thing I would say is align your social mission and your business model. Just was trying to say before that the social mission was a bit of a forcing function. We'll get that sorted out upfront. If you think of either as an-on, you will never be successful. It's like, oh, I am really a business, but I want to do this nice thing on the side, and it's never going to work. It's the opposite. You can't figure out a way to actually create business value, it will never work so you have to get that set up front. You need to make that call that says, "Okay, yeah, we are going to do this impact sourcing thing, but we're not going to be in a race to the bottom." You have to sort that and embed that into your strategy.

(01:24:54):
The second I would say is once you get that sorted out, you're seeking funding, you have to be upfront with who you are. You have to say that truly you are trying to do both. If you do not, you will get squeezed and pushed into decisions that you just don't want to make. And if you do not have the right investment in the right governance structure, your hosed. You're screwed.

Eric Ries (01:25:19):
Yeah. Get it right from the beginning. It's always too early until it's too late. That's one of my-

Wendy Gonzalez (01:25:23):
That's exactly right. And I'm so thankful. We went through so many months of figuring out what that structure was, and I'll also been times where I'm really thankful we made said no to an investor that we thought that was going to lead the round, because I can tell you, we would've ended up in the ground if we didn't get to that right investor. And it's really scary. You're like, "Oh, we're a tech company. We're all about SaaS multiples and all these things." And we're like, "Well, we have all these humans too. So yeah, we've got this awesome set of PhDs and software engineers, and we're so smart. We look like a hundred person startup. And oh, by the way, we've got 5,000 people in East Africa." But anyway, you don't really want to talk ... You have to be able to be upfront about that and say, actually, we are a unique business.

Eric Ries (01:26:11):
You can't be a defensive crouch about it. You got to wear it proudly on your sleeve, and you got to find investors who see it as a source of advantage. That is the only way to create the alignment that these-

Wendy Gonzalez (01:26:19):
And it took me a little while to figure that out. Honestly, we were almost a little apologetic about the mission. We're like, "Well, look at this rad technology. It's so awesome." And it took a while to realize that the only way you get there is by wearing it on your sleeve. So I would say that would be my next piece of advice.

Eric Ries (01:26:42):
Wendy Gonzales, CEO of Sama, thank you so much. Not just on behalf of all the people that you're employing and all the many people who you have inspired, but I want to say as a consumer of these products, thank you for being an advocate for their quality and for their safety, and for making them better and for creating this connection between the real craft of how they're made and the people that use them. Thanks for taking us behind the curtain there.

Wendy Gonzalez (01:27:07):
Absolutely. Thanks for having me on the show. I appreciate it.

Eric Ries (01:27:12):
You've been listening to the Eric Ries Show. Special thanks to the sponsors for this episode. DigitalOcean, Mercury and Neo4j. The Eric Ries Show is produced by Jordan Bornstein and Kiki Garthwaite. Research by Tom White and Melanie Rehak. Visual Design by Reform Collective. Title theme by DP Music. I'm your host, Eric Ries. Thanks for listening and watching. See you next time.

‍