Dive Deeper
Transcript
Episode Overview
AI guardrails for legal teams work best as runways. Each one is a short, per-function guide that gives one department a clear path to use AI without escalating every question to legal. That is how Holly Hogan handled AI adoption at Automattic while other GCs were telling their teams to wait for rules that did not exist.
Engineering received a two-page coding guide. Marketing received permissions mapped to existing IP rules. Consumer-facing teams received a disclosure framework built on consumer protection law. Then, she ran an internal campaign so people knew the runways existed.
Hogan earned the right to that playbook the long way. She spent nearly a decade as the legal backbone of Automattic, the company behind WordPress.com, Tumblr, and Day One, and steward of the open source software she describes as powering about 40% of the web.
She came in as the second lawyer in the building, built the team as the company grew from a few hundred people to roughly 2,000, and inherited a nine-year legal fight to protect a user in Turkey who had published content critical of the government. The Turkish government ordered the content removed. Automattic refused, and kept refusing.
Holly Hogan, General Counsel at Deepgram, said:
"It was nine years of really just standing behind, continuing to press, continuing to ask, continuing to push through there, and eventually got a victory, a really big victory. I mean, it was a long time coming, but being able to have a success against all odds, really sticking through that, I think that's something that's really unique from a legal perspective."
Cecilia Ziniti has known Hogan for a decade through the L-Suite community. In this conversation, they cover the guardrails Hogan built before AI usage policies were standard, why she believes lawyers should lead AI adoption, and the leadership habits behind a legal team colleagues called the best they had ever worked with.
About Holly Hogan
Holly Hogan is General Counsel at Deepgram, an enterprise voice AI company building toward the moment you can talk to an AI for five minutes without knowing it is not human. She joined in late 2025.
Before Deepgram, Hogan spent nearly a decade leading legal at Automattic, guiding the company through global litigation, privacy, payments law, AI strategy, and acquisitions. She then served as GC in residence at Mayer Brown, the first law firm to run that program, teaching practice groups what in-house work feels like from the inside. Before Automattic, she was an IP litigator focused on patent cases.
Key Takeaways
Standing still is the biggest AI risk for legal. While other GCs told their teams to wait for rules that did not exist, Hogan wrote approachable, department-specific AI guidance at Automattic instead of freezing until the law caught up.
Guardrails work best as per-function runways, not one policy. Hogan gave engineering a two-page coding guide, marketing permissions mapped to existing IP rules, and consumer-facing teams a disclosure framework built on the consumer-protection question of what a user needs to know.
Language shapes how the business treats legal. Hogan banned the phrase "the business" from her team's vocabulary, since legal is part of the business, and reframed legal work in business terms, saying "we helped close the deal" instead of "we did the legal review," until the mindset stuck.
Hire for expertise and curiosity in equal measure. When Automattic entered payments, Hogan hired someone with a fintech home base rather than a generalist, then paired that depth with people willing to grow into work they knew nothing about.
Clear prompting is clear thinking. Hogan argues the Socratic method, the interrogatory mindset of discovery, and a litigator's habit of layered questions all map directly onto working well with AI, which is why lawyers should lead AI adoption instead of resisting it.
What Are AI Guardrails for Legal Teams?
AI guardrails are the written permissions and boundaries a legal team gives each department so people can use AI without routing each decision through legal. Hogan organized hers by function rather than by legal subject matter. Instead of writing documents about privacy, IP, and confidentiality, she wrote guides shaped around how engineers, marketers, and consumer-facing teams would use AI day to day.
She has a name for the approach. Hogan explained:
"Let's get them guardrails. What I would describe as laying down runways. Let's build the asphalt, let's lay it down, let's build runways so they know where to go. They know the directions to take and how to move and get that plane rolling, and make it really simple."
The framing matters. A runway is infrastructure for motion, and Hogan's guardrails were built to get planes off the ground, with the limits marked clearly enough that nobody needed a tower clearance for each takeoff.
Why Standing Still Is the Worst AI Risk
Hogan's core position is that the biggest risk a legal team can take, when the rules do not exist yet, is refusing to move. Two years before this conversation, she was hearing GCs at events say they did not want their teams putting code into AI systems at all. They were waiting for rules that did not exist. She decided to do the opposite.
Hogan said:
"The worst risk here is standing still and we don't know what the rules are. I think being really open of saying we don't know what they are and that's okay. And what we're going to do is make the best judgments we have based on what we know. We have to shift, and that's fine."
Her reasoning goes beyond the legal team. Business colleagues worry about legal risk precisely because they cannot size it themselves. When legal projects hesitancy, that hesitancy trickles down through the whole company.
When legal says "we don't know yet, and here is our best judgment," the company keeps moving. Hogan calls it finding a path through the fog: make good decisions with what you know, and if you have to adjust later, the adjustments will land around the edges.
It echoes a theme from earlier in the season. Tina Patel, former VP and Associate General Counsel at Amazon Lab126 put it as "you got to make the call": take the information you have and make the judgment, because shying away from it is its own decision.
What Do Per-Function AI Guardrails Look Like in Practice?
At Automattic, per-function guardrails came in three flavors, each written with the team that would use it.
Engineering received a two-page coding guide developed with the engineering leads: how to use AI well, what to stay away from, kept clean and easy.
Marketing received practical permissions framed around what was different from existing IP rules, which was less than people feared.
Consumer-facing teams received a disclosure framework built on a first-principles question from consumer protection law.
Hogan explained the consumer-facing piece:
"Consumer protection law is always about what does the user need to know? ... And so translating that to the AI space, what are things that users need to know to be able to use this well that we should let them know about?"
Then came the step teams skip: distribution. Hogan ran an internal campaign, going team to team to put the guidance front and center so people could move with it. The result, in her telling, was a company that was forward in embracing AI at a time when she saw more hesitancy elsewhere.
If you are building the policy layer that sits above these guardrails, GC AI's guide to AI legal ethics covers the sanctions cases, the ABA and Florida Bar rules, and a traffic-light policy template you can adapt by department.
Why Lawyers Should Lead AI Adoption
Hogan argues that legal training is prompting training. The Socratic method in law school, the interrogatory mindset in discovery, the litigator's instinct to ask layered questions and iterate on the answers: all of it maps directly onto working with AI, which rewards a series of precise questions in dialogue.
Hogan said:
"Clear writing reflects clear thinking. I would venture to say that clear prompting reflects clear thinking too... lawyers should be some of the best adopters of AI in using it for work."
Her practical starting point is disarmingly simple: begin with the work you dislike and the work you love. She feeds uncomfortable emails to AI with a note about why they feel uncomfortable.
She uses it to wrangle PowerPoint layouts she has no patience for. And she uses it as an editor, what she calls a supercharged thesaurus, because her litigation days taught her she was a better editor than first drafter.
The lawyer who once took a brief to the coffee shop to tighten sentences now iterates drafts with AI instead.
For lawyers who want to build that prompting skill deliberately, GC AI’s Legal AI Classes teach it free, with California CLE credit, taught by former general counsels.
How Do You Build a Business-Forward Legal Team?
Hogan's philosophy comes down to two principles: pair lawyers with the business units they serve until they feel as much a part of those teams as they do of legal, and be deliberate about the language the team uses to describe its own work, because words build culture faster than almost anything else.
Hogan said:
"Don't say the business, we're gonna go talk to the business, like we are the business, don't say that... We cleared legal disclosures for this product launch? No, no. We helped launch the product. We helped our customers find clarity. We didn't do legal review for XYZ contracts, we helped close XYZ deals."
She reinforced that framing in group meetings, one-on-ones, and internal blogs until the mindset shift stuck, inside the team and across the company. Colleagues told her the result was the best legal team they had ever worked with, thought partners rather than a review desk.
Hire for Expertise and Curiosity in Equal Measure
When Automattic entered payments, Hogan knew nothing about payments law, so she hired someone with a fintech home base. That became the template. Hire deep expertise for the vertical, and hire the agility to grow beyond it.
"Having curiosity and agility and wanting to learn and grow into different things, that's the key, especially at a fast paced company. You're gonna need to hire the person who knows this or knows that, but you also want people that are willing to grow and jump into things that they don't know absolutely nothing about too."
The third ingredient is autonomy. Hogan gave her lawyers room to move inside their assigned business units, with a safety net when needed but without the second-guessing that teaches lawyers to play it safe rather than add value.
What Does a GC in Residence at a Law Firm Do?
As Mayer Brown's GC in residence, Hogan ran sessions with practice groups on positioning themselves for what clients need, held one-on-ones with partners and associates who sought her out, and worked on the firm's AI initiatives. Mayer Brown launched the program around 2023, modeled on the VC world's entrepreneur in residence, and was the first law firm to run one. Hogan describes it as immersive knowledge sharing.
The biggest value she delivered was a window most firms never get. She showed them what it feels like to bounce between employment law, a product launch, a media inquiry, an HR issue, and a regulatory matter, all before lunch.
"What they wanted was to talk a lot and work with them on just what's the unique challenges that we face as tech companies," Hogan said. "They're trying to be and are very AI forward, and very innovative in how they approach legal services."
What Does "Be an Original" Mean for Legal Leadership?
Asked for the one principle she wants listeners to take away, Hogan lands on originality. Build a legal function that could belong to no other company, led by a lawyer who could be no one else.
She said:
"Be an original. Think about originality... what makes you you, and be that... being who you are and leading into the best parts of yourself is really important. Don't try to be anybody else. But also from a meta level, at your company, what's unique and different about your company... How can you reflect those values?"
At an open source company, that meant radical transparency. At a voice AI company, it means building the legal infrastructure for a model racing toward a five-minute audio Turing test, in a field where the rules are still being written. That unwritten part is Hogan's favorite part. She also left the episode with one more reframe worth stealing: commercial lawyers are in sales, because the lawyers on the other side of the deal are an audience you have to move.
Hogan's advice is to find a path through the fog and keep moving. GC AI, the legal AI platform purpose-built for in-house counsel, gives your team the leverage to draft, review, and research at the speed the business moves, without adding headcount.
Recommended Reading
Why Judgment Is the New Superpower for In-House Lawyers in the Age of AI: extends Hogan's case for lawyers leading AI adoption, arguing judgment is what will separate legal teams as AI takes over routine drafting and research.
Generative AI for Legal: a month-by-month adoption plan that mirrors Hogan's per-function runways, rolling AI in one workflow and one team at a time instead of a single company-wide policy.
