Your Firm's Knowledge Is Trapped. A Chatbot Can Let It Out.
SharePoint didn't solve it. Intranets didn't solve it. A curated chatbot might.
Every firm has a knowledge problem, even if nobody calls it that.
Precedent briefs sit in a folder on somebody’s laptop. The client intake FAQ lives in an old Word doc. Half of what your firm knows — how you handle a discovery dispute, what you tell a new client about billing, which judge hates late filings — lives in people’s heads or buried email threads.
Big firms have poured money into fixing this for two decades. SharePoint sites. Intranets. Dedicated KM teams. The results have been disappointing. Solos and small firms mostly gave up before they started.
That’s starting to change.
What a curated chatbot is
A curated chatbot isn’t ChatGPT. It’s a chatbot that answers questions only from content you give it — your firm manual, FAQs, workflow notes, intake scripts, prior memos, whatever you feed it.
The mechanism has a name: RAG (Retrieval-Augmented Generation). When someone asks a question, the bot first retrieves the most relevant passages from your documents, then uses those passages to generate an answer. No open-internet guessing. Answers are grounded in your material, and a good RAG bot cites the source.
Think of it as an always-on assistant that has read every document in your firm and remembers all of it.
A working example
I run a community called The Inner Circle, hosted on Circle. Circle added an AI chatbot trained only on our community’s content — courses, recordings, posts, PDFs, two years of discussions.
Members use it constantly. Real questions they’ve asked:
“Tell me about Lawmatics.”
“Can you show me a Clio workflow when a client contacts the office?”
“Summarize Ross Fishman’s presentation on niche marketing.”
“What’s the name of that service that clips web pages, Audible clips, and Kindle bookmarks?”
When the bot misses, I jump in and add my take. I’ve become the backup to the bot, not the other way around.
The side benefit surprised me. Because members ask the bot instead of silently giving up, I get a window into what’s on their minds — what they’re working on, what they’re stuck on, what they care about. Low-friction for them. High-signal for me.
Why this matters for small firms
For twenty years, real knowledge management has been a big-firm problem with a big-firm budget. The results were mostly bad, but at least big firms could afford to try.
Curated chatbots flip the economics. A solo with a folder of documents and a free NotebookLM account can put together something surprisingly useful in a weekend.
New hires ramp up faster. The intake team gets instant answers to the same twelve questions clients ask every week. You stop being the bottleneck for questions you’ve already answered a thousand times.
How to start: NotebookLM as training wheels
Don’t buy enterprise software. Don’t hire a consultant. Start with NotebookLM, Google’s free RAG tool.
NotebookLM is nothing more than a RAG with a clean interface. You pull sources in — PDFs, Word docs, Google Docs, web pages, YouTube videos — and it answers questions using only those sources. No open-internet wandering. Every answer points back to the source it came from.
That’s the whole thing. Which is why it’s the right starting point. It forces you to think like a RAG builder: curate the sources, ask the questions, watch what breaks.
Your first project should be small:
1. A pile of your firm’s content — manuals, FAQs, prior memos, intake scripts, anything text-based.
2. A short list of real questions your team or clients ask every week.
3. A weekend to iterate.
One note on what to put in: this isn’t the place for raw client data. Start with firm procedures, intake scripts, redacted precedent memos, and onboarding materials. That covers most of what your team and clients keep asking. (Putting actual client data into a chatbot is a whole other article.)
Bottom line
Your firm already has the knowledge. It’s been trapped in PDFs, email threads, and people’s heads for years.
A curated chatbot finally makes that knowledge useful — without a big-firm budget and without a KM department. Pick one document pile. Feed it to NotebookLM. See what happens.
;-)
Ernie
P.S. In the Inner Circle, we’ve got a chatbot trained on two years of AI tool discussions, courses, and member contributions — so you don’t have to dig through hundreds of posts to find an answer.
→
https://innercircle.ernietheattorney.net/