Aliisa Rosenthal, OpenAI's first sales leader, has found a new career in venture capital. She joins Acrew Capital as a general partner, working alongside founding partner Lauren Kolodny and the firm's other partners, Rosenthal and Kolodny told TechCrunch.
Rosenthal left OpenAI about eight months ago after a three-year sprint at the AI Lab where he launched products such as DALL·E, ChatGPT, ChatGPT Enterprise, and Sora. “I had no intention of joining a VC fund initially,” she told TechCrunch. “I was meeting with a lot of AI startups.”
But after growing OpenAI's enterprise sales team from two people to several hundred people, she realized the appeal when she received a venture capital offer from Kolodny. Instead of helping one startup with its go-to-market strategy, she might be able to help with a portfolio of those companies.
During her time at OpenAI, she said, “I learned a lot about buyer-side behavior, how people think about these purchases, and the gap between what most organizations think is possible and what they can actually deploy today.”
For example, she has first-hand insight into what kind of moats AI startups can build so they don't leave themselves vulnerable when model makers like OpenAI launch competing products.
Is OpenAI “just going to build everything and put all companies out of business? You know, they're already doing a lot of things: building consumer, enterprise, devices. I don't think they're going to pursue all the potential enterprise applications,” she says.
So there is one moat for enterprise AI startups to offer their expertise.
tech crunch event
San Francisco | October 13-15, 2026
Context as a moat
Additionally, she believes that the key to a good startup moat is “context,” or the information that the AI stores in context window memory as it acts on a request.
“Context is dynamic. It's adaptable. It's scalable. And I think what we're seeing is moving beyond some kind of basic RAG toward the idea of a persistent context graph,” she says, referring to search augmentation generation (RAG), which is the de facto way in 2025 to minimize hallucinations by training the LLM with specific sources you trust (and having the LLM cite them).
However, there are still many technologies that need to be developed in this area, from memory to reasoning beyond pattern recognition.
“I expect real innovation here. I think we'll see a new approach this year: the idea of context and memory,” Rosenthal says.
But beyond startups working directly on context engineering, Rosenthal believes there are benefits for enterprise apps that incorporate context engineering.
“Ultimately, when we talk about moats, I think who owns and manages this context layer is a huge advantage for AI products,” she says.
Another opportunity she sees is startups that aren't building cutting-edge models from big labs at high prices.
“I think there is room in the market for cheaper models that are lighter weight and have revolutionized inference costs,” she says. These models probably won't be at the top of the leaderboards of various benchmarks, but they are “still very useful” and more affordable.
“What I'm really excited about investing in is the application layer. I'm really interested in what durable applications will look like built on all of these different models, not just the basic model,” she says. She's looking for startups with “interesting use cases” or that use AI to help companies work more efficiently.
As for where she will find these startups, she plans to start by building a network among OpenAI alumni. Now, ten years after the AI organization was founded, its alumni network is growing. Many companies have already founded startups and raised big money at high valuations, from OpenAI's biggest competitor Anthropic to buzzy early-stage companies like Safe Superintelligence.
There are also increasing instances of former high-level OpenAI personnel becoming seed-stage investors. About a year ago, Peter Deng, former head of consumer products at OpenAI, joined Felicis. Since then, he's been crushing it, landing big deals with hot startups like LMArena and Periodic Labs, and clearly having fun.
“I actually had a call with Peter a few months ago and he helped me make the decision,” Rosenthal said of his choice to become an investor.
But Rosenthal may have a secret weapon to winning a contract. She also has deep connections with AI enterprise users, the type of buyers and beta testers that early AI startups need.
Companies still don't realize how much AI can do for them. “There's a huge gap and I'm very optimistic that we can close it. There's a huge greenfield left for applications and enterprises.”