Inside Liberty's AI Adventure: Lessons & Laughs

Liberty Mutual's AI Odyssey: A Roller Coaster Ride with Lessons and Laughs

Alright folks, buckle up because we're diving into the wild world of Generative AI (GenAI) deployment at Liberty Mutual. Picture this: a Fortune 100 company with a tech landscape as varied as a TikTok FYP, from legacy C++ systems to AWS cloud wonders. Liberty Mutual decided to ride the AI wave, and let's just say, it wasn't all smooth sailing.

Garth Gilmour and Stuart Greenlees, the dynamic duo from Liberty IT, recently shared their journey at the GOTO Copenhagen 2025 conference. Spoiler alert: it was a mix of triumphs, face-palms, and a lot of learning on the fly.

Lesson #1: The AI Hype Train

Remember Thanksgiving 2022? While we were busy stuffing ourselves, OpenAI dropped ChatGPT, and the tech world collectively lost its mind. Liberty Mutual, like many, was caught off guard. Gilmour likened the experience to surfing a wave: "You got to swim up to the top...if you feel yourself going off, you have to keep paddling to get back up."

The takeaway? Always be ready to pivot. Technologies can go from 'interesting' to 'essential' faster than you can say "AI takeover." Keep your research game strong, and don't underestimate the power of timing.

Lesson #2: Shipping ≠ Adoption

So, you've rolled out a shiny new tool. Congrats! But spoiler: that doesn’t mean everyone’s using it effectively. Liberty learned this the hard way. They launched LibertyGPT to 70% of their team, but actual usage? Meh.

The solution? Embrace the chaos. Liberty's teams started having spontaneous "unconferences" and "inspirational sessions"—a.k.a. peer-to-peer learning with a side of dad jokes. Gilmour says, "It’s easy to inject humor...you can get it to talk like a pirate." Arr matey, indeed.

Lesson #3: Models Aren't One-Size-Fits-All

Here's a plot twist: not all AI models are created equal. Liberty Mutual found that different models excel at different tasks. Need jargon-heavy responses? OpenAI's your buddy. Want simplified language? Try Gemini models.

Currently, Liberty is juggling 28 models. Yes, 28! Talk about commitment issues. But hey, when it comes to AI, it's all about finding the right tool for the job.

Liberty's AI Adventure: A Few More Bumps

The journey to AI integration is paved with tech hurdles and learning curves. From technical glitches to regional regulations, Liberty's path hasn't been easy. But with each "swing and miss," they’ve found a way to recover and hit it out of the park.

Liberty's Experiment in AI-Powered Chaos

Deploying GenAI at scale is like trying to teach your grandma how to use Instagram filters—frustrating, hilarious, and ultimately rewarding. Liberty Mutual’s experience shows that embracing mistakes, fostering a culture of spontaneous learning, and customizing your AI toolkit are key.

So, next time you’re faced with a tech challenge, channel your inner Liberty: embrace the chaos, keep learning, and maybe, just maybe, talk like a pirate. No cap.

---

By Zara Chen, Tech & Politics Correspondent at Buzzrag