In January 2022, Gro Intelligence's CEO, Sara Menker joined Gro investor Hans Tung of GGV Capital and Emily Ma of Google at Stanford School of Engineering’s Entrepreneurial Thought Leaders series. They walked through how Gro Intelligence is using AI, data, and human intelligence to model complex real-world systems across climate and agriculture and help organizations see around the corner with more confidence. This session has been cut for length and clarity.
See the full session here
Building a Company to Address Huge Problems at Scale
Emily Ma (Stanford, Google): One of the most pressing challenges of our time - maybe throughout the next 10,20,30 years - is really climate. Climate issues are requiring us to work together a little differently, in a cross-sector manner. Public, private, academic, nonprofit, and all of these different entities need to work together and find a path forward.
Sara, you came out of Wall Street as a commodities trader. You knew that you wanted to get back to solving for food insecurity, and you had a number of options. How did you choose to go the route of a VC funded startup?
Sara Menker (Gro Intelligence): I didn't go down the route of a VC funded startup. I went the route of solving a problem that I was very passionate about. Solving for that problem led me to understand that what I needed to do is create a company, not a nonprofit, for example. There's also that route.
Starting with a problem that you're genuinely passionate about often leads you to the right solution. To me, the first realization was just how complex the challenge was going to be and how much funding it would need over time:
The different interests you'd have to weave
Farms that are half acres and farms that are 100,000 acres
Very small food and beverage companies and mega companies
You have to deal with a very complex system.
As a business, you can sort of design yourself around how do I solve for that problem? And so, we started by really saying, who are the domain experts that we wanted to hire, as opposed to the technologists we're going to hire?
Because the domain experts then said, "Here's how I think…" When we hired our first environmental scientist, it was because I saw their resume and I said, "Oh, I've never worked with satellite imagery before but this sounds like something we should do. We should hire this person."
That deep domain expertise, forming it, and then bringing in world class engineering to scale that human knowledge became the lens by which we took that path. And I think one of the great things about taking the path of a VC-funded startup - or it's the path of a business - is that you innovate and you're super agile and you change your ways as you get more data and more information than you have.
I always say that you have to be driven by your mission and your purpose. But the strategy, and especially the tactical moves, you take from a day-to-day, month-to-month basis, they will change - while that final end goal always remains the same. And I think that works really well for the types of investors that we brought on board early, which is what VC is great at.
Human Intelligence Drives Artificial Intelligence
Emily Ma (Stanford, Google): It's unique that you started not with the tech but with the domain expertise because, ultimately, we have to be applicable to the domain experts at the end of the day. Let me ask a deeper question.
What sorts of organizations tend to be your direct customers? How do you partner with them? How do you share your products with them? And how do you communicate with other stakeholders in the system who may not necessarily be your direct customers, if that exists too?
Sara Menker (Gro Intelligence): So very good question. We have a saying at Gro which is, "The AI is only as good as the HI behind it." So you better get your human intelligence right. Because otherwise, you won't build a very smart system. You might build an AI that is just not very good.
And so if you care to solve for it, start with that. This is fundamentally an important lesson for AI companies in general, whatever sort of domains we're talking about. You have to make sure that the team resembles the world you're trying to model whether it's in the domains that they're from or the languages they speak or the way that their minds operate.
This notion of diversity expands beyond, "Are you a man or a woman? Are you black?" It is, "Do you have a little microcosm of the world you're modeling sitting in your company?" If you don't, you're gonna have other challenges down the line. So be really smart about that. That's what has allowed us to think about the supply chain end to end.
If you think of an agricultural system - and I call it the anatomy of agriculture - you have people who feed into the system. What products should you develop? What are people gonna want? And for where are you developing - because that's hyper local. And from a policy standpoint, how do you manage national food security risks, from strategic reserves, management?
Food security is national security in most countries.
Then you go further down. We don't really work with farmers because farmers are hyper optimized to a field and we optimized the system. We really think of sort of the whole world in sort of an interconnected way. But once the farmer grows that product, it needs to go somewhere. So you have logistics companies and you have trading companies that actually trade or processing companies that process that product. And so there's applications there because, ultimately, they want to know how much was produced and of what type. And so the same models that you've built at the top actually impact the buying side.
And then you have one more layer which is the companies that brand the goods. So a CPG company that creates chocolate bars or creates your soups, or whatever that is because all of those are the ingredients that go into it. And ultimately your retailers that sell it.
Essentially we touch everyone across the system except for the farmer and the end buyer, like you and I. We're not a consumer product. And we're not a product for the farmer to optimize like the precision farming type role.
But there are many other players in the ecosystem, as you rightfully note, that you need to get in. You have lots of advocacy organizations, you have lots of universities, academic research institutes, that generate tons of research that we leverage and scale sometimes. We're not reinventing the wheel everywhere.
Recently, we started a program called Gro For Good, where members work on a problem that's going to solve a challenge in a net positive way for society. You can get access to any level of our product for free.
I'm really excited, especially this year, about bringing Gro For Good to life in a much bigger way because I think that if we are going to be sort of a trusted source for the industry, we have to be trusted by everyone. That's I think a really important principle.
Emily Ma (Stanford, Google): That's beautiful. I wanted to go back to something that you said earlier that really struck me and reinforced this notion. To get the AI right you have to get the HI right - you know the human intelligence right. And again, and again, you bring that up. A couple of hours ago today, I was talking to a chef who said, “With all your computer vision work, there's a difference between parsley stems and cilantro stems.” And I'm like really? They look exactly the same. But apparently he can tell, just by seeing them. You need the domain experts in the room to ground truth and to guide the development of products. Otherwise, we're not going to get it right.
Hans, switching over to you. You saw me around the corner pretty early on too and have developed this deep expertise and interest in food tech and ag tech. You've invested in a number of companies in this area, obviously with Gro, as others, and as one of the top VCs in the world, you have an opportunity to invest in literally anything. What brought you into the space? Why are you so excited about it? How did you find your way to Sara? Tell us more about this.
Hans Tung (GGV Capital): I've discovered my career in investment banking. So with Sara, I worked at Morgan Stanley, I was doing the investment banking side, at Merrill Lynch on Wall Street. That led me from New York to Hong Kong. And then a group of friends from Ivy Leagues, Stanford, and Berkeley wanted to do a startup. I ended up doing two startups. Both of them didn't make any money, but I learned a lot.
I ended up coming back to join a VC firm and started doing VC both here in the US, as well as in Asia. And the advantage of having three years of banking experience, three years of growth capital investing experience, and then four years of startup experience was that I was more ready to take advantage of opportunities.
I thought that world food security would be a bigger problem. Given climate change, it will be increasingly harder and harder to grow food efficiently and effectively. And at the same time, food wastage was becoming a bigger problem, as well. A lot of the things that we are sold into the grocery stores, a lot of them end up being wasted and perishable items.
The question we ask ourselves - can we find great founders that we can back that can solve this systematically? Learn the lesson before: acquire the right domain knowledge. At the same time, approach it from a sort of tech lens, from some of the software companies build. Smart tech companies that know how to scale and know how to create and iterate really quickly and efficiently. So not everything is moving in a matter of months, but matter of days or weeks. So have an internal clock that can fast turn around and iterate, give feedback, and then innovate again.
When we invested in Bowery Farming, we asked - who else do you know in New York or other places that you have come in contact with solving the problem with a very promising team? That was Sara and Gro intelligence.
We look for founders who are doing amazing work, and then that person's network would lead us to other promising founders who think alike, have similar kind of approaches to solving problems. Even the company may have different backgrounds. That's what makes the job so interesting.
Pivots vs. Incremental Changes
Emily Ma (Stanford, Google): Every business over time evolves and pivots. We talked about iteration earlier. We could take the pandemic, for example. It affected any and every business on earth. Was there a period during Gro’s history that was challenging for whatever reason that basically caused you to pivot and adapt in a unique way?
Sara Menker (Gro Intelligence): There are many challenges. And I think it goes back to a point Hans made, though, which is you need to make these incremental changes really, really fast, not wait for sort of massive pivots to occur to redefine who you are. And that is the difference between what I mentioned around mission and vision versus the tactics to get there.
Have we changed the path by which we're getting to our end goal? Absolutely. And we constantly continue to fine tune that path, whether it's determining where exactly to sell our product, how to reach our audience, how to present ourselves to the world and speak about ourselves.
Last year, one of the big things we did was a complete rebranding. Our logo used to be circled in red, and now we’re blue, and it has a completely new look. And that involved a massive journey. It's not a pivot, but it's actually a seminal change in the way that we then interact with the world. Thinking about it as small incremental changes, that's all my job is. To find out and fine tune what we're doing constantly. And sometimes, if you're actually a person that's super structured and you'd like to have complete clarity on what to do, a startup is disrupting and disruptive all the time.
Real World Impact, Gro For Good, and Fighting Locusts
Emily Ma (Stanford, Google): I want to talk about a few more questions if possible. Sara, what kind of real world impact have you seen through the distribution of the Gro platform? Particularly Gro For Good, which may be quite new right now or otherwise?
Sara Menker (Gro Intelligence): Yeah. One is impact that you make to the operation of an organization or the decisions they make, and seeing them drive change in their businesses as a result of suggestions or outputs of your model. That, as I said, is the ROI path. And that one's actually, I think, the easiest.
The second is where you actively go out to help solve for some type of challenge, and then you're either successful or not, and really, Gro For Good came about from a few years ago.
We started a data science team at Gro called the Rapid Response Data Science team. And the goal of the Rapid Response Data Science team was to be the ER of data science. So if something really unprecedented happened, and we needed to build a model super fast, not through the full engineering lifecycle, the prototype model, put it out there, see if it sees record floods in a region that has never experienced it before. And we built this team to do that so that we can put things out to the market, as fast as possible to get that feedback and see how well the models perform.
Well, at the beginning of COVID, another crisis that was happening was the locusts crisis across East Africa going into South Asia. And the spread of locusts was threatening the lives of hundreds of millions of people. And, with everybody distracted with COVID, there were not enough resources that were being deployed and not enough money being raised for fighting and battling the locusts.
And so our Rapid Response team, plus actually a huge part of the company, volunteered over a few days and built out a Locust Impact Kit. We open sourced the models to basically assess both the spread of and the potential impact of locusts on crops in every single country that locusts had been detected.
And so we scaled a methodology that required flying planes to see if there were locusts to then looking at a combination of imagery and applied our knowledge of how crop yields work to all the different regions we've done. We developed yield models, plus assessments of how bad the locust impact would be, to then do an assessment of how many dollars need to be raised. Versus what had been being raised by the UN, which at the time was $350 million. We knew very quickly over $3 billion was needed.
In opening that up, we got about 43 public organizations around the table. We were on 2am phone calls with governments trying to help them wake up to how bad this would be, to mobilize that.
Gro For Good came really from that experience saying, we did a lot of good. We used the resources we had, at a time, frankly, when the company itself was trying to figure out what COVID meant for us. But it was our role to do that, right? And so Gro For Good was: how do we use our platform to let many others do that because we as a company can't step in for every crisis considering the pace at which we're facing crises. And the cost of these crises is going up. There are going to be many countries and many use cases where they're not necessarily commercial. Gro For Good exists because of one of these projects that we've done.