Dive Deeper
Transcript
Episode Overview
Jimmy Toy opens this episode with a dare: replace me. The Chief Legal Officer of Articore Group, the parent company of Redbubble and TeePublic, has spent nearly 12 years defending two of the world's largest artist marketplaces, and he now spends his energy teaching AI to do the parts of his job that no longer need him.
His method gives in-house counsel a direct answer to a question more legal teams are asking. How do you build a legal knowledge base with AI? Feed it the litigation record.
Toy's team loaded a decade of court filings, affidavits, declarations, and deposition transcripts into an AI folder his lawyers can question like a colleague, and the cross-jurisdiction consistency checks that once cost him three hours of searching now come back in minutes with the exact passages he verifies against the source documents.
About Jimmy Toy
Jimmy Toy is the Chief Legal Officer at Articore Group, the publicly traded parent company of Redbubble and TeePublic, online marketplaces where independent artists sell their designs on print-on-demand products. He joined Redbubble in 2014 as a corporate generalist, back when the company had never been sued, and became Chief Legal Officer in 2022.
Over nearly 12 years he has managed 40 to 50 IP cases across the United States, Australia, Europe, and Asia, guided the restructuring of Redbubble into the Articore Group parent structure, and supported the launch of creator storefront platform Dashery and the May 2026 acquisition of India-based creator marketplace Frankly Wearing. He has used AI and machine learning in his legal work for more than 10 years, predating modern LLMs, and speaks on platform liability and risk at the Marketplace Risk conference.
Key Takeaways
Lawyer resistance to AI is a trained identity problem, not a skill gap. The billable hour taught lawyers to prove value through hours logged, so compressing a thousand-hour work product into five hours threatens the business model as much as it threatens the lawyer's self-image.
An LLM can already give executives 80% of a legal answer. Toy says a lawyer's remaining value lives in the other 20%, the judgment calls, institutional context, and years of trust an AI cannot replicate.
A legal knowledge base starts with the documents nobody rereads. Toy's team loaded a decade of court filings, briefs, affidavits, declarations, and deposition transcripts into an AI folder that lawyers can query like a colleague.
Three standing values keep a legal team steady through growth, contraction, and turnaround. Toy's team runs on light touch, no surprises, and making every dollar count, since legal's clearest lever on the board's numbers is operating expense.
The "replace me" system turns AI delegation into a five-step process. Identify the repetitive work, hand it to AI or a teammate, build workflows and escalation paths, measure the results, and keep iterating.
Why Are Lawyers Resistant to AI?
Toy's diagnosis is that the resistance is trained into the profession. Lawyers learn to demonstrate value through flawless independent judgment backed by visible time and effort, and the billable hour converted that identity into an economic engine. Compress a thousand-hour work product into five hours and the output stays as good, but the business model and the professional self-image both wobble.
Toy described the psychology:
"As a lawyer, you feel like you're responsible for everything, every output, every response... I think it does make it harder to embrace a new tool that is new and risky and uncertain, but has such potential, I think, to transform the way we do our work."
Going in-house loosens the grip. Toy says he learned to value the timeliness of his advice over the hours behind it. But a subtler conflict survives the move. In-house lawyers build standing with executives by being the person with the answers, and AI now competes for that role.
Toy put the uncomfortable math plainly:
"You want to show that you're the person that they can come to that has the answers... But actually, they could have gone to the LLM and asked and probably gotten 80% of the answer that I gave."
That question frames the rest of the episode: what lives in the other 20 percent?
How Do You Decide What Legal Work to Delegate to AI?
Start from the replacement question. Toy inverts the fear of obsolescence into a design exercise: list what you do repeatedly, what you dislike doing, and what an AI system or a teammate could do instead, then build a system around the split. The system includes workflows, policies, escalation paths, and measurement, and it improves through iteration.
"I kind of turned the 'oh no, we're gonna get replaced' mindset on its head and say, replace me, how can I be replaced?"
Delegation targets range wider than the legal team. The work can go to a paralegal, a junior lawyer, or a self-help resource that lets the people and culture team answer their own questions without waiting on counsel. Toy's five-year vision decentralizes the GC's monopoly on answers:
"Create these self-help tools for people on other teams to just kind of replace you as the person they had to ask before and wait, because you're a bottleneck and you're expensive. It gives them an alternative. It's a lot cheaper for the company and a lot faster."
He backs the philosophy with management structure. Articore's legal ops manager is the team's AI champion, and Toy gave her a standing assignment that pushed past prompt-writing:
"Every week I want you to walk me through a tool that you created... And I don't just mean a prompt that... generates a good output. I'm talking about something that people can use to self-help... that's built on our knowledge base."
How Do You Build a Legal Knowledge Base With AI?
Collect the entire litigation record, load it into an AI-accessible folder, and interact with it like a person. Toy's version holds court filings, briefs, affidavits, declarations, and deposition transcripts from 40 to 50 IP cases spanning nearly 12 years. The archive can live in something as simple as a shared drive folder connected to an LLM, or in a purpose-built projects feature inside an AI platform.
Toy explained why the comprehensiveness matters:
"You may just focus on your briefs, like the big briefs that you filed, but here you can really put it all in and create this excellent knowledge base that's comprehensive. And then you can just interact with it like it's a person."
The payoff shows up in consistency. A platform defendant across dozens of jurisdictions has described its business under oath in France, Australia, the United States, and India, and a contradiction between filings hands opposing counsel a gift. That consistency check used to run on Toy's memory alone.
"It would fall on me to try to make sure everything was consistent. And then I might spend three hours trying to figure that out and maybe not come to an actual answer on it. But now I can just not even do that... I can just farm it out to the AI tool, and of course you go back and check... but you never would have been able to locate those and find the exact passages without that sort of help."
The same base cuts onboarding costs when a new case lands in a new jurisdiction. New outside counsel gets up to speed on years of arguments without billing the archaeology hours. For teams building this workflow, Files in GC AI, the legal AI platform built for in-house counsel, holds permanent document collections accessible across chats and analyzes up to 1,500 pages at once, while Projects carries matter context between conversations. GC AI's guide to the best legal AI tools for in-house counsel compares the platform options by fit.
What Does a Decade of IP Litigation Teach About Risk Management?
Losing sometimes is evidence the risk calibration is right. Redbubble had never been sued when Toy joined in 2014; within six months the first case arrived, and 40 to 50 IP matters followed across the United States, Australia, Europe, and Asia. His scorecard philosophy:
"I like to say we win most of them, but we don't win them all... I think if we won all of them hands down, knocked it out of the ballpark, we might be managing risk with too heavy a hand."
The case he is proudest of started with an outlaw motorcycle club. Hells Angels sued Redbubble in Australia for trademark infringement after users uploaded designs featuring club logos, and Australian law at the time had a copyright safe harbor comparable to the DMCA but no settled framework for trademarks on user-generated content platforms.
Redbubble lost at trial. The appeal, [Redbubble Ltd v Hells Angels Motorcycle Corporation (Australia) Pty Ltd [2024] FCAFC 15](https://www.wipo.int/wipolex/en/judgments/details/2791), split 3:2, with the majority narrowing the injunction into a notice-and-takedown safe harbor rather than eliminating it, setting aside the \$70,000 additional-damages award, and substituting \$100 in nominal damages.
"I like to say I testified against the Hells Angels in federal court... So we actually lost our first round... this case went on 10 years. And then we just in the last couple of years won on appeal. And I'm very proud of it because it actually created law in Australia."
The decision to keep appealing was a balancing exercise: costs, likelihood of winning, disruption to the business, and settlement leverage.
Fellow guest Michael Jacobs, the former Morrison and Foerster partner now at JAMS, made the same point on his own CZ and Friends episode: deciding to bring a litigation is like deciding to open a business unit. Toy agrees from the defense side. A class action pulls in the CTO for depositions, engineering for discovery, marketing, and finance, and it has to be staffed, paid for, and managed like any other company-wide initiative.
How Does Legal Add Value in a Turnaround?
Through OpEx. Articore has moved through growth, contraction, and now a turnaround phase focused on getting the stock price up, and Toy is direct about what changes. Legal cannot make the business grow, but it has a direct line to operating expense, and operating expense is profit. A legal team that reduces outside counsel spend moves a number the board watches.
Three values hold the team steady across cycles, and they predate the turnaround:
1. Light touch: Advise without slowing the business down.
2. No surprises: The board cannot control risks it never hears about, so legal's job is informing the people who need to ask questions in time to ask them.
3. Make every dollar count: Cost discipline as a standing habit, so a downturn requires no culture change.
"We keep trying to add things to that list of three, but... that list of three is pretty much everything. Everything kind of goes back to it."
What Should Legal Leaders Tell Their Early-Career Selves?
Toy's lightning-round answers compress the whole episode into three lines: stop fearing imperfection, measure output over hours, and trust systems over heroics.
"Always try to rely on a system and build a system and continuously improve your system... It makes it so much less stressful because you don't feel like you're starting from scratch every time."
His book pick follows the same logic. Beyond the sci-fi he reads for fun, the text that shaped him as an executive is the workbook companion to Peter Drucker's The Effective Executive, a fill-in-the-blanks volume he still completes exercises from years later, because writing the answers forces the application.
And the thing he wishes he could shortcut: none of it. Asked what surprised him across nearly 12 years, Toy names the relationships, the CEOs and CFOs he has watched come and go, and the trust that compounds into what he calls a great legal product for the company. AI gets the archive, and the 20 percent stays his.
Recommended Reading
Replace Me: Articore Group CLO Jimmy Toy on Using AI to Make Himself Obsolete: GC AI's companion article expanding on this episode's replace-me framework.
AI in Legal Operations: The 2026 Five-Layer Playbook: the framework behind the delegation-and-self-help system Toy describes.
In-House Legal Bench: Evaluating AI Assistants for In-House Legal Work: how GC AI benchmarks AI performance on the legal tasks Toy hands to a knowledge base.
