The Lawyer Who Gave the Legal Department a Reply Button

The most overworked person in a company is often the one everybody asks whether they are allowed to do something. Can marketing say that? Can sales sign this? Can we hire there? The in-house lawyer lives amid a pleasant fiction: that the legal department is a department. In practice, it is an inbox with professional liability.

GC AI was built around that inbox. It researches law, reads and compares documents, drafts clauses, checks contracts against a company's preferred positions, summarizes policies and carries its work into Microsoft Word. By September 2026, the company said more than 2,100 legal teams in 53 countries used the platform. Named customers have included News Corp, Nextdoor, SKIMS, Liquid Death, Vercel, TIME, Zscaler, Eventbrite and Tipalti.

The lawyer employed by the client

Cecilia Ziniti had already been general counsel or chief legal officer three times - at Anki, BloomTech and Replit - before she co-founded GC AI with Bardia Pourvakil in 2023. Pourvakil had worked at Replit and Roam Research. One founder knew why the business wanted an answer before lunch; the other knew how software communities adopt new tools. Their distinction was simple enough to sound obvious only after somebody said it: an in-house lawyer and a law-firm lawyer do not have the same job.

The law firm sells careful legal work. Corporate counsel must turn that work into a decision amid product launches, budgets, personalities and a risk tolerance that changes with the deal. A research database may know the law. A general chatbot may produce a polished paragraph. Neither necessarily knows that this company accepts a twelve-month liability cap, hates automatic renewals and needs a plain-English answer for a sales vice president in six minutes.

Ziniti first experimented with pre-ChatGPT language models on contract drafting and review in 2022. She and Pourvakil then started a newsletter about AI and legal work. Education was not a decorative marketing layer; lawyers needed to know what the machine could do, where it could fail and how to ask it better questions before they would trust it with sensitive work.

The first thing legal AI broke was not the billable hour. It was the illusion that fluent prose and dependable advice are the same thing.

A machine with receipts

The earliest constraint was trust. An invented case citation is comic until it appears under a lawyer's name. GC AI says Ziniti personally reviewed early responses and pushed the product toward a standard a practicing lawyer could defend. That instinct survives in Exact Quote, which ties extracted language back to a document, and in research answers designed to point toward primary legal sources.

The platform uses models from outside providers, including OpenAI and Anthropic, while GC AI supplies the legal system, workflow and verification around them. It says those providers do not train on customer data, and its trust center lists SOC 2 Type II, SOC 3 and GDPR controls. Enterprise plans add the usual gates demanded by security teams: single sign-on, directory sync, domain verification and audit logs.

This is also where GC AI separates itself from both ends of the market. Against ChatGPT or Claude, it sells legal context and controls. Against Harvey, Legora and other legal platforms, it emphasizes in-house work, a Word-centered routine and transparent individual pricing. Against Westlaw and Lexis, it offers a conversational workbench that drafts and edits as well as retrieves. The crowded market now includes all of them - plus Google and Microsoft moving legal agents into the productivity suite. GC AI's moat cannot merely be access to a good model. Models travel. Habits, company playbooks and trusted workflow travel more slowly.

The arithmetic of ten hours

The individual plan costs $500 a month, or $5,000 paid annually. Team and enterprise prices depend on scope and seats. That is dear beside a consumer chatbot and modest beside outside counsel. GC AI makes the comparison explicit: at $475 an hour, the subscription pays for itself in fewer than ten avoided hours. Its December 2025 customer study reported an average 14 hours saved per lawyer each week and a 14 percent reduction in outside-counsel spend.

Those are company-reported figures, not neutral commandments. Yet they describe the right buying test. Do not ask whether an AI can write a memo. Give it last month's real work - the messy contract, the unfamiliar employment question, the board draft - and count the time from prompt to reviewed work product. The expensive answer is sometimes the one that appears quickly and takes an hour to verify.

GC AI's own benchmark tested 100 in-house tasks against attorney-built answer keys. The company reported an 86.8 percent pass rate, seven points above the nearest general assistant tested. The methodology is more interesting than the victory lap: specific tasks, explicit criteria, source documents and human spot-checks. Any buyer can copy that architecture with a dozen of its own completed matters.

From newsletter to legal operating layer

The money arrived after unusually fast reported growth. At the November 2025 Series B, GC AI said it had moved from $1 million to $10 million in revenue in under a year; investor Northzone described roughly $11 million in revenue, driven largely by word of mouth. Scale Venture Partners and Northzone co-led the $60 million round, with Sound Ventures, Aglae Ventures, SilverCircle Partners, News Corp, The Council and strategic investors participating.

By the summer of 2026, the company was shipping like a business trying to outrun platform convergence: an API, Easy Edit, Agent Connectors and Contract Intelligence arrived in a concentrated launch week. It also won the 2026 SaaS Award for highest customer satisfaction. The cultural machinery is published in six principles, including “ship today,” “customer obsession” and “own it.” The charmingly non-automated detail is that Ziniti reportedly spends about half her week talking with customers.

The human stays on the signature line

GC AI works best where there is enough repeated work to encode and enough legal judgment to make acceleration valuable: a lean department reviewing familiar agreements, researching across jurisdictions, updating policies or answering recurring business questions. It is less persuasive for an occasional user whose needs fit a cheaper general tool, or for a team unwilling to build playbooks, test outputs and change its routine. Bad source documents remain bad source documents. A contract repository full of drafts, duplicates and missing amendments will not become wise merely because it acquires a chat box.

Nor does the product remove professional responsibility. GC AI's terms require appropriate human oversight. Novel litigation strategy, a bet-the-company investigation or advice turning on facts nobody supplied still belongs to experienced counsel. The platform can compress the first pass, surface the clause and organize the questions. It cannot be the person whose name appears beneath the answer.

That may be why the company is interesting beyond legal technology. GC AI did not win attention by pretending expertise had become free. It treated expertise as an expensive, interrupted human resource and built a machine to return some of it. The reply button is not the answer. It is the chance for the lawyer to spend less time composing one and more time deciding whether it is right.

A practical note: Legal AI is a work tool, not legal advice. Pricing, customer counts, benchmark results and savings cited here are current public or company-reported figures as of September 2026 and may change.