Your inbox already knows the pattern. The CRM announces an assistant. The portfolio platform announces an assistant. The planning tool, the document vault, the custodian: each one, this year or last, unveiled AI inside its product. None of them are lying about it. Each assistant is real, each one works, and each one can see exactly one thing: that vendor’s own system.
The boundary is the product
Your CRM’s AI can read what lives in your CRM: contacts, notes, tasks, the emails it syncs. Ask it to summarize your history with a household and it does a decent job. Ask it what that household holds, and it has no idea, because holdings live in your portfolio system and the custodian’s feed, and the CRM has never seen them. The portfolio platform’s assistant is the mirror image: fluent about positions and performance, blank about the conversation your advisor had with the client in May.
Every one of these assistants inherits the borders of the system it ships inside. That is not sloppiness. It is the shape of the deal.
Why this is their business model, not a flaw
A vendor’s AI exists to make that vendor’s product more valuable and harder to leave. It is a retention feature, and a good one. Building the assistant your firm actually needs, one that reads across the CRM, the portfolio system, and the custodian together, would require each vendor to treat its own database as just one source among several, including its competitors’. No vendor is paid to do that. Their moat is the data inside their walls, and an assistant that reached across the walls would be an assistant that made the walls matter less.
So expecting your CRM vendor to solve cross-system AI is expecting them to fund the end of their own lock-in. They are not villains for declining. They are just not the ones who will ever build it.
The test you can run today
You do not have to take any of this on faith. Open whichever built-in assistant your firm already pays for and ask it one question that crosses systems. A good one:
- Which of my clients over seventy hold the fund we put on the watch list, and which of them haven’t had a review meeting since January?
Ages and meetings live in the CRM. Holdings live everywhere but the CRM. Then watch what the assistant does, because there are only three moves available to it. It can say plainly that it cannot see the other system, which is the honest answer. It can answer the half it can see and hope you notice. Or, worst of all, it can answer confidently as if its half were the whole, and hand you a client list with the most important column missing.
If that question does not fit your firm, build your own from the same recipe: take one fact from one system and one fact from another, and ask for the intersection. Which households mentioned a move in a meeting note this year, and what state is their mailing address in now? Who opened an account in the last quarter, and has anyone scheduled their first review? Any pairing works, as long as the two halves live behind different logins.
Run the test on each assistant you have. You will learn more about your stack in ten minutes than a year of vendor webinars will teach you.
What answering it actually takes
The cross-system question is answerable. It just has to be answered from outside any one vendor’s walls: the firm’s records, drawn from the CRM and the portfolio data and the custodian, standing together in a database the firm owns. That last part is the part that matters. When the database is yours, no vendor can end the arrangement by closing a door, and the AI that reads it works for the firm rather than for a product.
That is the shape RIAI is built on: one AI across the whole firm’s records, a receipt on every number pointing back to its source, scope enforced in code so each person sees only what they should, and an append-only record of all of it. Your vendors’ assistants will keep getting better inside their walls, and you should use them there. The work between the walls was always going to be somebody else’s job.