OpenAI sales leader says ‘we see ourselves as AGI sherpas’

For OpenAI head of sales Aliisa Rosenthal, there is no disconnect between the company’s lofty mission to “ensure that artificial general intelligence (AGI) benefits all of humanity,” and her team’s day-to-day sales, marketing, partnership and customer success efforts, including implementing ChatGPT Enterprise at over 260 companies and 150,000 employee users.

In fact, in a recent interview with VentureBeat, Rosenthal, who reports to OpenAI COO Brad Lightcap, described her now 150-member go-to-market team as “AGI sherpas” who “help our customers and our users transition to the paradigm shift of AGI” — even though there is no agreed-upon definition of AGI across the AI industry. She also made no secret of the fact that OpenAI uses feedback from customers to improve its models towards AGI, and that the original goal of ChatGPT was simply to gather more data for GPT-3.5.

The OpenAI sales and GTM (go-to-market) team, which Rosenthal said does not work on quotas or commission, continues to grow quickly — the company’s Careers page shows over two dozen roles available. For Rosenthal, who says she took a risk joining OpenAI two years ago after a four-year stint as VP of sales at SaaS company WalkMe — months before ChatGPT led to the current generative AI boom — discovering the power of GPT and DALL-E was “magical.” She said she knew there must be a business application and was “pretty determined to find out what that was.”

In the interview, which has been edited and condensed for clarity, Rosenthal also discussed her first chat with OpenAI CEO Sam Altman; her father’s prescient prediction about OpenAI; and her team’s sherpa emoji.

Sharon Goldman: You recently posted on LinkedIn saying that the OpenAI sales team had started out as a ‘tiny rag-tag team.’ Now you have over 100 salespeople. How crazy was that arc of evolution for you?

Aliisa Rosenthal: When I joined two years ago it was a really small go-to-market organization, probably 15 people or so. When ChatGPT launched, we were probably up to 30 people. You can imagine we had a lot of interest in ChatGPT, so we basically spent the last year growing the team as fast as we could. We’re now almost 150 people. 

Where have the moments been on this journey where you were like, wow, I have to take a minute and calm down?

When I started here, AI was really relegated to experimental groups, beta projects, and small startups. What’s been really fascinating about ChatGPT is even though it is a consumer product, it ignited the imagination of boardrooms across the world. Suddenly it felt like every executive was thinking, “How do we start implementing AI in our business processes and the products that we build?” It was an unexpected shift that was obviously a pivotal moment where we thought, “Okay, it’s time to really invest in and build our ability to support these enterprises as they’re asking these questions.” 

The second big moment came when we released our ChatGPT enterprise product. We have a real product market fit with this and it’s been really exciting to launch and begin to build a customer success organization around it. We launched in early September and then very quickly acquired hundreds of companies as enterprise customers. That was another tipping point on the scaling adventure.

What was your first conversation with CEO Sam Altman like? I read somewhere that he said something like, “I don’t even know what we’re selling.” What was that like for you coming from previous roles? 

I’ve been in tech for almost 20 years now. I’ve worked at a bunch of fast-paced startups. And I worked at an AI startup about a decade ago, which gave me a little bit of an introduction to neural networks and cluster analysis and a little bit more appreciation than I might have otherwise had when I first tried out GPT-3, which was December 2021. 

I had reached out to OpenAI — I had a friend working here and he introduced me to Sam, who was very candid, saying “I don’t exactly know what sales is going to look like here. It probably won’t look like anywhere else. It will probably be a little bit different.” 

I was so impressed with the audacity of his mission and the goals of the organization. I took a big risk. I actually joke that when I accepted the offer here all of my venture capital friends told me not to take this role. They said to just go somewhere with product market fit, where you have a big team and everything’s established and figured out. Why do you want to go to this research lab with barely any product? The one person who told me to take the job was my dad, so he gets a lot of credit for that. 

Why did he feel strongly about that?

My dad’s a mathematician and had been following LLMs in AI and OpenAI, which I didn’t even know about until I called him and told him that I had a job offer here. And he said to me — I’ll never forget this because it was so prescient— “Your daughter will tell her grandkids that her mom worked at OpenAI.” He said that to me two years ago. 

That’s crazy!

I know. I was like, “Dad, you knew more than all my venture capitalist friends.” So I’ll give him some credit. He’s been an inspiration to me and my biggest cheerleader. 

I didn’t know you had reached out to OpenAI. What was it that your friend was telling you that sparked your interest? 

The CRO at the company I was at previously, WalkMe, became president of Jasper, which was one of OpenAI’s original big customers. He was recruiting me to Jasper to run sales. I was digging into documentation and I saw GPT-3 and kind of went down the rabbit hole I found on LinkedIn. I knew a couple of people at OpenAI and thought, let me ping them to see what they think of Jasper. That led to one of them saying well, don’t join the application, come join the platform. I asked, are you guys building a sales organization? And he said, ‘I think so. Let me introduce you to Sam.’  So it played out in a funny twist of fate. 

Did you envision the OpenAI products that could come down the pike?

I think companies have been trying to solve chatbots forever, so that seemed like the most obvious use case we could figure out. I also saw DALL-E when I was interviewing. At the time, I wasn’t totally sure how you’d sell that to an organization. But it’s like once you see that you can’t unsee it. Once I saw the power of GPT-3 and the power of DALL-E I just began thinking everything is going to change. I can’t just go work for a traditional company now that I’ve seen the technology around the corner. 

I wouldn’t give myself that much credit to say I saw everything that was coming, but I will say I was very spellbound by the magic of the models. As part of my interview process, they asked me to build an app using Python. It’s a very technical organization and they wanted me to really grasp the power of the API and the model.  I made a little bot that basically was a translation app. It was just totally magical for me. 

Did you feel that Sam had good instincts as far as productizing OpenAI’s models?

When I joined, honestly, I wasn’t 100% sure that we would find the right product market fit. I thought, okay, well, this could just be a really interesting experiment where I spend a year dabbling in AI and learn a lot about it and then go on my way and find a more traditional SaaS company. I did think at the time it was just that it felt so revolutionary. It felt so magical. I just thought there’s got to be some sort of business application for this. And I was pretty determined to figure out what that was.

When did you really see that ChatGPT was “the thing?” 

I think we all underestimated the impact it would have because we’d been playing with those models in-house for so long. At that point, we were all just focused on GPT-4, that was what was really exciting for us. ChatGPT was a way to get some more data on GPT-3.5 and some feedback. It was honestly a side project, it was not what we were all really focused on.  But we realized pretty quickly once we released it that it was going to have an outsized impact that we had not prepared for. 

One thing about OpenAI that I’ve struggled with is understanding its dual mission. The main mission is building AGI to benefit all of humanity, and then there is the product side, which feels different because it’s about current, specific use cases. 

I hear you. We are a very unique sales team. So we are not on quotas, we are not on commission, which I know blows a lot of people’s minds. We’re very aligned with the mission which is broad distribution of benefits of safe AGI. What this means is we actually see ourselves in the go-to-market team as the AGI sherpas — we actually have an emoji we use  — and we are here to help our customers and our users transition to the paradigm shift of AGI. Revenue is certainly something we care about and our goal is to drive revenue. But that’s not our only goal. Our goal is also to help bring our customers along this journey and get feedback from our customers to improve our research, to improve our models. 

But when you talk about AGI with an enterprise company, how are you describing what that is and how they would benefit from it? 

One is improving their internal processes. That is more than just making employees more efficient, but it’s really rethinking the way that we perform work and sort of becoming the intelligence layer that powers innovation, creation or collaboration. The second thing is helping companies build great products for their end users, whether that’s building a brand new support solution to help their customers find answers, or a new way of shopping, or a new experience with their customers.

But what I mean is, how do you connect that to the larger, long-term mission of OpenAI? The AGI mission seems very long-term and futuristic to me.

You’re right, AGI is not happening tomorrow. But there are step functions on the way to achieving AGI where we can really supercharge innovation, we can have a really big impact on organizations with the tools that we have today. So an example is our customer Moderna. They rolled out our ChatGPT enterprise products to all of their employees. In particular, Moderna is leveraging our advanced data analysis tool to help them run analysis on dosage data, which they found decreases the drug approval process for them by 30 days on average. The way they put it, when you are a child waiting for life saving treatment, 30 days can be the difference between life or death. So to me the Moderna story is a step in the direction towards AGI. It’s improving the way that we all innovate and create and frankly makes the world a better place. 

So you’re seeing it as a step-by-step process as opposed to talking to Moderna and saying, 50 years from when we have AGI you guys can do such and such. 

This is where the sherpa emoji comes in. It’s helpful to progress in this journey rather than suddenly have AGI dropped on you out of the sky. It’s helpful to start to understand AI and integrate it into your processes and workflows. Similarly for us, it’s really helpful for us to get feedback on how companies are using it, how users are using it. If you think of AGI, the definition is autonomous systems that can perform work as well as humans. We have to understand how humans work. So I think it goes both ways. It’s healthy for us in our mission of creating AGI to understand and get feedback from our users. And then it’s good for our users and for organizations to begin to integrate AI into their practices.

And you’re still hiring people  — how many more people do you need? 

We are hiring like crazy, as fast as we can, while keeping our talent bar high. We’re looking for people who are not coin-operated, but who are more invested in successful customers, fascinated about AI, interested and optimistic about the benefits, more thoughtful about the risks. We’re looking for people in every single go-to-market function, who are really driven by our mission.

How does the sales team handle the OpenAI news of the day when you’re meeting with potential customers? Just last week, for example, there were issues with ChatGPT going off the rails. 

I don’t know what you’re referring to.

People were posting examples of ChatGPT talking in Spanglish and giving nonsensical answers. 

I’m not up to date on that. There’s so much to stay up to date on here. I don’t really know enough about that particular issue.

Can you talk more about your priorities going forward in 2024? 

I think this year is really all about moving from hype to reality, from experimentation to implementation, and that’s going to be hard. I think we have to have a thoughtful strategy. We have to be able to help our customers embrace change. It’s hard to depart from existing norms and a more traditional way of doing things. This is where the idea of a sherpa is really going to come in. 

Similarly, it’s about moving from a sales organization that is used to supporting tech companies and startups to supporting Fortune 500 companies. We’re working with more traditional industries, so we are also maturing as an organization, building things like event marketing and a customer advisory board, and really understanding the needs of large enterprises. 

Some studies that I’ve read say that organizations are slow to adopt some generative AI use cases. Do you see that changing, as far as moving from the most internal, low-risk use cases to something that really would be put out into production to customers?

I very much agree with you on that. I think in the beginning it took companies a while to really understand how to take this and implement it and I think there were so many possibilities. A lot of boardrooms and executive teams were almost paralyzed by the decisions ahead. I think that we’re starting to break through that and organizations are starting to experiment, try things and get it into production. 

I think Coca Cola is an interesting example. They basically use GPT-4 and DALL-E 3 to create this sort of AI platform where consumers can come and use the creative assets from the Coca-Cola. It was just such a magical way, I think, for a company to leverage AI to expose their end users to a completely unique experience. 

But the Coca-Cola example is about creative. Hallucinations could actually be good in that kind of situation. Are you concerned that some use cases that companies are chomping at the bit for can’t be implemented until issues like hallucinations or safety issues are improved?

I think the models will continue to get better and better, will have better reasoning capabilities, who even knows what this year is going to bring in terms of model capability — but I think we will see this space continue to evolve very, very quickly and possibilities grow ever wider for what companies are comfortable simply deploying. I do think we’re seeing a lot of organizations embrace this for internal employee facing use cases, where you’re not necessarily you know, having the model operate independently, but your employees are leveraging it, to boost their own productivity and to come up with new ideas. It’s really interesting. We’re finding ChatGPT is helping them actually improve the output of their work, not just do their work faster. 

What is it like working with COO Brad Lightcap?

Brad is somebody who really trusts the people around him and lets them run their franchise. He’s not a micromanager. He’s very open to ideas, to suggestions, to being challenged. So it’s been really fun and rewarding working with him. I also think one thing I really appreciate about Brad is he’s not somebody who has one strong opinion that cannot be changed. He’s somebody who is flexible in his thinking and he’ll always say go try it, experiment and see what works and then come back to come back to me and we’ll take it from there. I think that is part of the OpenAI culture too that I really love, which is you very rarely get told no. If you have an idea, you want to do something, are usually told, put a proposal together, run an experiment, and if it works, we’ll go for it. That is very much part of the culture of the whole organization.

Have there been any experiments that you’ve run as a sales organization? 

Constantly! Endlessly! One example is we created a new role in July called Account Associate, which I didn’t even totally know what it was going to be, but we have had all this inbound, a really big challenge of responding to people reaching out to us. I wanted to build a team, not necessarily a BDR team because I wanted them to also be able to answer technical questions from customers — more than just a traditional business development rep who’s just trying to qualify sales leads. 

Now we have a team of 10, soon to be 20 of them. It’s sort of the front line, we created a really great experience. So the first time you reach out to us, you can talk to a person who knows our products and can sort of direct you in the right way. That’s sort of a unique role that I don’t know if it exists anywhere else. 

Do you feel confident that your OpenAI sales sherpas will make it to the top of the mountain?

I mean, we’ve got a big mountain ahead. I’m not saying it’s all going to be easy. We have a lot of hiring to do. You know, these spaces move quickly, and we need to stay on top of it and understand all the different applications. We need to help our customers navigate it. We have to help them with the cultural transformation required to be successful with AI. But I’m really optimistic and I’m really excited. With the success we’ve seen so far with some of our customers, I’m really excited to see more companies embrace this and see similar successes.