Rachel Shields Williams really knows her stuff! From a brief conversation with Doug Striker, CEO of Savvy Training, to her keynote speeches, professional life and work, and this particular podcast, her thoughts and commentary are very insightful and relevant.
Law firms are spending real money on AI right now, and a lot of that spend rests on a quiet hope: that the new tool will finally make sense of decades of documents, precedents, and matter files nobody ever organized. It won’t. Not on its own.
That’s the core message from a recent episode of Adventures in Legal Tech, where host Jared Correia talks with Rachel Shields Williams, Director of AI and Knowledge Management at Bass, Berry & Sims and President of the Legal Marketing Association. Her point is one we see every week in our own training rooms: firms kept everything, but almost no one structured it. AI doesn’t fix that. It exposes it.
Here are the ideas from that conversation we think every firm should sit with, plus what we’d add from the training side.
AI can’t organize what only humans understand
Rachel compares most firms’ data to an overstuffed closet: everything is in there, and opening the door is a risk. Her blunt test for the “AI will sort it out” pitch is that if it were true, she’d already be retired.
The reason is context. A firm’s data reflects its client relationships and its intellectual property. Knowing which brief was the winning version, which clause is the firm standard, or which matter is confidential takes human judgment. Structuring that is people work, and it must come first.
She offers a useful frame: think of AI as the middle layer between two things you control.
- Below it: your structured data — clean, tagged, findable.
- Above it: your firm’s expertise — the judgment, playbooks, and know-how that make you different.
- The middle: the large language model, whose pricing, features, and availability you don’t control.
If you invest in the two layers you own, you can swap the middle one as the market shifts. If you skip them, no model will save you.
Shadow AI is bigger than a chatbot
Most firms define shadow AI as a lawyer pasting client data into a public chatbot. That risk is real, but Rachel widens the lens. Open-source code pulled from places like GitHub can carry hidden backdoors from bad actors, and it often slips past the review a new vendor would get.
Her analogy is a good one for any firm policy: you hand a niece a safety knife, not a chef’s knife. The goal isn’t to ban the tools. It’s to give people a safe version to learn on, then widen access as skills and guardrails mature.
This is exactly where security awareness and AI training overlap. People need to know not just how to use AI, but which tools are approved, what data can go where, and how to spot something that doesn’t belong.
Small steps beat big launches
Rachel’s change-management math is simple: get 1% better most weeks of the year and you end up roughly 48% better, without blowing up anyone’s workflow. (If those gains compound, it’s closer to 60%.) That beats a firmwide “AI day” that everyone forgets by Friday.
She also likens using technology in a firm to swimming in the ocean — you have to respect the riptide. Our take: the current is the pace of change in AI itself. Fight it head-on and you tire out. Work with it — steady, deliberate progress — and you get where you’re going.
For training, this argues for short, repeatable, role-specific learning over one-time sessions. People retain what they use the following week.
The rise of the translator
One of the most useful ideas in the episode is a role Rachel sees emerging in law firms: the translator. This is the person who can move between lawyers, IT, and firm leadership. They explain why data structure matters, what AI can realistically deliver, and what it can’t.
We’d go a step further. Every firm needs translators, but it also needs a workforce that can meet them halfway. A translator can’t carry adoption alone if attorneys and staff don’t understand the basics of how AI tools use firm data.
That’s the gap we focus on at Savvy Training & Consulting. Our AI courses for Copilot and Claude are built around a firm’s own policies and guardrails, so people learn the approved tools, the right use cases, and the limits — in the context of their actual work.
Five questions to ask before your next AI rollout
- Is the data ready? Do you know where your best work product lives, and is it tagged well enough for a tool to find it?
- Who owns structure? Is someone accountable for organizing knowledge, or is it everyone’s side job?
- What’s your safety knife? Which approved tools can people experiment with safely, and how do they request more?
- Who are your translators? Who bridges the lawyers, IT, and leadership — and do they have the time and authority to do it?
- How will people learn? Is training ongoing and role-specific, or a one-time launch event?
The bottom line
AI is only as good as the knowledge underneath it and the people using it. Organize the closet, give people a safe place to start, and build skills a little every week. That’s how AI moves from expensive experiment to everyday advantage.
If your firm is planning an AI rollout — or cleaning up after one — let’s talk about training that fits your policies, your tools, and your people.
Sources
- ‘Adventures In Legal Tech’: Where Knowledge Management Meets AI — Above the Law, October 13, 2026
- Mind Map – The Intersection of Artificial Intelligence & Knowledge Management — Adventures in Legal Tech, full episode
- Full episode on YouTube