Editor’s note: The following is the fifth profile in a series of articles about startup founders from non-technical backgrounds who have launched successful venture-backed companies. Read the previous interviews with MagicSchool AI founder Adeel Khan here, Trunk Tools founder Sarah Buchner here, Tabs founder Ali Hussain here, and Craft founder Ilya Levtov here.
When an executive at Replit suggested that Cecilia Ziniti seemed distracted by artificial intelligence, she was not an engineer working on the coding startup’s AI products. She was its general counsel.
The executive told her ‘You’re a great GC and good at business development … but it seems like maybe your head’s not in it because you’re so into this AI thing,’” Ziniti recalled in an interview with Crunchbase News.
“And it was true,” she said.
By then, Ziniti had spent roughly two decades working as a lawyer at companies such as Amazon, Cruise and Anki. She had never learned to code. She had not ever founded a company. And she had not worked in the same sort of technical or product roles that many venture-backed software founders have.
But her time at Replit gave her an early look at generative AI. That was enough to inspire her to leave the legal profession and build GC AI, a startup bringing AI technology to legal work.
An early look at GPT
In late 2021 and early 2022, Replit was working with OpenAI on AI-powered coding products, giving Ziniti access to an early version of GPT before ChatGPT was released publicly. She began considering what the technology could mean for lawyers.
Ziniti later taught classes on using ChatGPT for legal work. During these sessions, she began to see the disconnect between applying a general-purpose chatbot to a profession that requires precision, sourcing and tone.
For instance, in one class, she used ChatGPT to research the legal considerations involved in entering the Brazilian market. The response covered the relevant issues, but it started out with language about going to Rio and grabbing surfboards, Ziniti recalled.
At that point, Ziniti recognized there was a distinct gap between an AI-generated answer and something a lawyer could actually use in a professional setting. She realized that what broad models weren’t quite getting right could — and should — become product requirements.
As she began developing the idea, Ziniti teamed up with Bardia Pourvakil, an engineer she had worked with at Replit who had experience with earlier generations of GPT. Pourvakil had once been admitted to Stanford Law School, she said, but chose engineering instead, partly because he expected AI to automate some legal work.
“I would teach these classes in the morning and tell my co-founder in the afternoon, ‘Our software needs to provide accurate citations. Our software needs to tell you where in the document it’s getting the backup for this. Our software needs to speak in this way,’ ” she said.
A founder who was also the customer
Their backgrounds complemented each other. Pourvakil knew how to build the technology, while Ziniti intimately knew the work it was meant to support.
“I am a better founder for GC AI than I think anyone could be because I was the ICP,” Ziniti said, referring to the startup term “ideal customer profile.” “I was able to step up and meet the moment, even though I don’t code.”
Her understanding of that customer came from a legal career that began well before the current AI boom. Ziniti started as a paralegal at Yahoo, later worked at Morrison & Foerster and went on to hold senior legal positions at Amazon, Cruise, robotics startup Anki, and Replit.
Although it was when Ziniti realized the enormous potential of AI when it came to the legal industry, Replit wasn’t her first experience with the technology. In 2013, she became the first full-time lawyer assigned to Amazon’s Alexa. She later encountered other forms of AI through autonomous vehicles at Cruise and robotics at Anki.
Notably, she had done much of her work inside companies rather than at law firms. With that experience in mind, Ziniti decided to focus GC AI on those in-house legal departments.
From idea to company
Ziniti left Replit on Nov. 1, 2023, and incorporated San Mateo, California-based GC AI about a week later.
The company initially built an AI assistant for corporate legal departments, then added products for contract analysis and for handling requests submitted to legal teams.
GC AI has seen impressive growth, growing 400% year over year, according to Ziniti. The company now serves roughly 2,100 companies, up from about 900 a year earlier, she said. Its customers range from large enterprises to startups with a single in-house lawyer, and include companies such as Lockheed Martin, Time, Eventbrite, Vercel and Gusto.
The company makes money through a mix of per-seat subscriptions and additional products, including a contract-analysis tool priced by document volume. It also sells a platform for handling and responding to requests submitted to legal departments.
While GC AI competes with legal AI startups Harvey and Legora, Ziniti said the startups mostly serve different customers. Harvey and Legora initially focused on law firms, while GC AI has concentrated from the beginning on corporations and their in-house teams.
Its contract intelligence offering, for example, is designed to analyze a company’s own documents, while law firms’ tools may need to work across materials belonging to thousands of clients. Nearly 30% of GC AI’s seats are also used by employees outside legal departments, including HR and finance executives, according to Ziniti.
She believes her extensive legal background helped GC AI address one of the biggest barriers to selling AI software to corporate legal departments: trust.
“It is the number one most important thing,” she said.
About one-third of GC AI’s 125 employees are lawyers, according to Ziniti. The company pursued SOC 2 compliance early, built data-isolation protections, and says it does not train its models on customers’ confidential information.
One longtime customer, Ziniti recalled, described their perception of the difference between GC AI and general-purpose AI tools this way: “Trust, trust, trust. Oh, and trust.”
A different route into venture capital
Ziniti was new to founding a company, but not entirely new to the world of venture capital. She had invested as an angel and served as general counsel at several venture-backed startups, which helped build investor relationships before she began raising money herself.
One of GC AI’s first commitments came from Amber Illig, founder of The Council, a venture firm that invests in experienced operators becoming first-time founders. Ziniti had previously worked with Illig at Cruise.
“Once you get one commit, then it’s relatively easy,” Ziniti said of the company’s seed round.
GC AI has since raised nearly $72 million across three rounds. Its most recent financing was a $60 million Series B co-led by Scale Venture Partners and Northzone, which valued the company at $555 million. Sound Ventures, Aglaé Ventures, SilverCircle Partners, News Corp and The Council also participated.
Ziniti said about 45 general counsels have also invested in GC AI through a special-purpose vehicle.
Her experience illustrates one way generative AI is producing a different type of software founder. One who is an experienced industry practitioner who has identified a use for a certain technology but needs a technical partner to build it.
In Ziniti’s case, the idea for GC AI emerged from the overlap between two parts of her career. Replit exposed her to generative AI early, while her years as an in-house lawyer gave her firsthand knowledge of what a legal AI product should do.
“You have a unique insight on the world in some way from your life experience,” she said. “I did, and I built around that.”
Related Crunchbase query:
Related reading:
- ‘Nobody Wanted to Give A Former Principal Money’: How An Educator Built An Edtech AI Startup With $63M From VCs
- How A Teenage Carpenter Became The Founder Of AI Construction Startup Trunk Tools
- From Humanities To AI: How Ali Hussain Built Fintech Tabs Into A $400M Startup
- Craft’s Ilya Levtov Never Learned To Code. He Built A Software Company Anyway.
Illustration: Dom Guzman
