On the name. AUM is how this industry keeps score. AI is the other thing worth putting under management: pointed at the right work it does a great deal, and left ungoverned it becomes a problem you didn’t budget for. Both halves matter, and most firms are only thinking about one.
This issue: three things you can build this week, the one thing that’s probably blocking the fourth, and two stories worth your time.
Three You Can Build This Week
None of these touch a client record, so none of them wait on the compliance conversation below.
1. The money-in-motion brief
Clients show up when money moves. A business sells. A lockup expires. Someone retires. A plan terminates. Almost all of it gets filed publicly, and almost nobody watches it.
Build a weekly brief off public filings and regional business journals. Form 144 is the one worth knowing: insiders file it before they sell, so it runs ahead of the trade. Add 8-Ks for acquisitions and executive departures, and your local business journal, which is often faster than any filing.
Score what comes back by size, recency, and how close it is to you. Take five. For each one you want a dated source, a reason it matters this week, and a first line you’d actually send. Then run it against your CRM so it never hands you someone you already know.
Everything in it is public. Your CRM is only used to subtract.
One honest caveat. By the time something is filed you’re often late. The real value is calibration: it shows you which attorneys, CPAs and bankers keep turning up next to the money, and those are the relationships worth building now.
2. The quarterly letter
Every firm writes one. Same shape every quarter, same reader, and it still gets built from scratch.
Teach it your format, your voice, and last quarter’s letter. Give it this quarter’s numbers and the market context. You get a draft, plus a list of what actually changed since last time.
No client data required. Performance at the strategy or model level is firm information, not client information.
You still write the view. It writes the version around it.
3. Meeting prep
You have a first meeting Friday with someone you met at an event. Right now you’ll skim their LinkedIn in the car.
Instead, hand over the name and the company and get a page back before you leave. What the business does, how it makes money, what’s happened there in the last year, who else sits at the table, and three questions worth asking that they haven’t been asked twenty times already.
It’s a prospect, not a client, and everything in it is public. Nothing from your book goes near it.
Fifteen minutes of reading you didn’t have to do, and it shows in the first ten minutes of the meeting.
The Memo
The Permission Slip
It goes the same way every time I’ve seen it.
A few people at a firm start using AI on their own. Nobody planned it. Within a few weeks they’re getting hours back every week, and the work coming out is better than what it replaced.
So naturally they want to go further. Stop pasting things in by hand and let it reach the systems where the work actually lives. The CRM. The portfolio system. The accounting file.
That’s when someone asks the question in a meeting. Are we allowed to do that? Nobody knows. And once it’s asked out loud, you can’t un-ask it.
Everything stops. Not because compliance said no, but because nobody had asked yet, and now they have to. I’ve watched that pause run a full quarter.
It isn’t a technology problem. It’s a plumbing problem, and it’s a smaller job than it feels like.
Start with the Claude plan you’re on
From here this is Claude specific, and that’s deliberate. No other major assistant gives you a compliance API at all, which is the difference between a tool you can show an examiner and one you can’t. If you’re picking an AI for a registered firm, that one fact should decide it.
So check which Claude plan your firm is actually on. Most firms get this part wrong and it takes thirty seconds.
Free, Pro or Max. That’s a personal account. Your firm has no administrator, no access, and no contract with the vendor. If client information is going in there you have two problems: you can’t produce those conversations, and since June 3 you’re also out of step with Regulation S-P, which now requires you to actually oversee the vendors holding client data. You don’t have a vendor to oversee. You have an employee with a subscription.
Team. Better, and it fools people. An owner can export conversations. But it’s a manual export, run by one person, and any employee can delete their own chat first. There’s no audit log. That’s retrieval, not capture, and the difference shows up the day somebody asks you to prove something.
Enterprise. This is where it starts working. You get audit logs and a compliance API, which is the piece an archiving vendor can pull from continuously, without anyone remembering to.
Before you budget
Enterprise on its own isn’t the answer. The compliance API doesn’t store anything. It makes the record available, and you still need an archiving vendor on the other end to hold it.
Anthropic already keeps the data long enough, six years for compliance sessions, which clears the five-year books and records requirement. So what you’re buying isn’t storage. It’s a record nobody can quietly alter, and proof somebody reviewed it.
There’s also a change in how you get billed, and it catches finance teams off guard. Team is a flat subscription. Enterprise is a seat fee plus usage, so the number moves with how much your people actually use it.
Two things control that. Set per-user spend caps low and raise them for whoever hits them, because raising a cap is easier than clawing one back. And set a firm default model, since which model people reach for drives the bill far more than headcount does.
Meanwhile, don’t stop
Two tracks, and it’s tempting to run them in sequence. Don’t.
Start the Enterprise conversation now, and put the archiving contract on its own track, because contracts are slow and the upgrade doesn’t have to wait on one.
Then keep building. The line is simpler than people make it: a specific client’s account, holdings or AUM stays out until this is sorted. Your market research, investment process, marketing and prospecting do not. That’s most of what’s above.
One last thing, for the day you flip the switch. Once you’re on Enterprise with an archive behind it, the firm can retrieve any session. That’s the point of it. It also means the chat window stops being a private notebook, and your people should hear that from you rather than work it out later. Treat it like the company email account. Same rules, same expectations.
It’s worth being clear about what’s on the other side, because that’s the actual reason to bother.
Once client data can go in, the work changes shape. Meeting prep stops being a page about a company and becomes everything you’ve discussed with that client and what’s moved since you last spoke, ready before the call. Review packets assemble themselves out of the accounts instead of getting rebuilt every quarter. You can put a question to the whole book at once, who’s overweight this, who hasn’t heard from us in six months, and have it back in a minute instead of an afternoon.
None of that is exotic. It’s the same tool you’re already using, pointed at the data it currently isn’t allowed to see.
The compliance work isn’t the project. It’s the permission slip. Go get it, and build while you wait.
What Moved
Claude’s compliance API now covers Cowork and Claude Code
Anthropic extended its compliance API on August 11 to cover Claude Cowork on desktop, web and mobile, and Claude Code. Enterprise only, and still in beta. It doesn’t reach Claude Code on the web, or anything running on Bedrock, Vertex or Foundry.
My take: this is the agentic side of Claude becoming usable at firms like yours, and that’s where most of the leverage is. Don’t rewrite your AI policy around a beta. Do put it on the agenda for your next compliance review, and ask your archiving vendor when they plan to support it.
63% of RIAs use AI. About one in ten have built anything on it.
Schwab surveyed 533 advisors who custody on its platform. AI use has more than doubled since 2023, to 63%. Of those, 82% are using generative AI, mostly through individual experimentation rather than anything the firm set up. Only about one in ten are fully integrating it into how the business runs.
My take: that gap is the opportunity, not the indictment. Almost every firm already has someone who’s good at this. What they don’t have is a governed place to do it and a reason to build something that lasts. Sort the second and the first compounds fast.
That’s it for the first issue. The next one lands in two weeks.
If it was useful, forward it to whoever at your firm is going to get handed the AI question. Usually that’s the COO, or the person who inherited compliance along with three other jobs.
And if you want to talk about what this looks like at your firm, in your workflows, on your timeline, just reply. That’s what we’re here for.
Tim
COO, Bridge AI Consulting
bridgeaiconsulting.com
This is general information about published rules, not legal or compliance advice, and nothing here creates an advisory or attorney-client relationship. Run your own facts past your own counsel and CCO.
Tim built the internal AI workflows and training systems that got developers at Epic Systems using AI every day. He's a Partner and COO at Bridge AI.
