Your CRM Is Ready for AI. Your Data Probably Isn't.

Here's something we've seen happen consistently as advisors start using AI inside their CRM: the first thing AI does isn't answer a question. It surfaces one.

An advisor asks their AI to segment their book — show me all clients above a certain asset threshold, or all clients in a particular life stage. The answer comes back wrong. Or incomplete. Or it doesn't come back at all.

The advisor assumes the AI isn't working. What's actually happening is that the same type of client has been categorized three different ways in the database — three different naming conventions, accumulated over years of different advisors and inconsistent data entry. The AI sees all three versions and can't reconcile them.
The CRM wasn't the problem. The data was.

The System of Record That Isn't

Most firms think of their CRM as their system of record. What AI reveals, pretty quickly, is that "system of record" often means "place where records live" — not "place where records are consistent." Whether the data is organized in a way that's actually usable is a different question, and most firms haven't had to answer it until now.

What we've learned building Redtail Assistant is that data quality is the real constraint for AI in advisory firms. Not the tool. Not integration. The data underneath it.

This isn't unusual — it's structural. CRMs accumulate years of input from multiple people with different habits. Notes get written at different levels of detail. Clients get categorized differently across time and staff. Nobody cleaned it up because nobody needed to. AI needs to.

What Good CRM Data Actually Looks Like

When people ask me what a CRM needs to look like before AI is genuinely useful, I think about two things.

The first is history — a long record of interactions. Emails sent, activities completed, workflows that ran. That longitudinal record gives AI context over time. The second is richness — detailed notes that capture not just what happened, but why. A note that records a distribution from a client's account doesn't tell the whole story. What were the tax consequences? Which accounts did it come from? What were the relevant considerations at the time? That context is what shapes future conversations in ways that matter.

Most CRMs I've seen have history. Richness is hit or miss. But at minimum, the data needs to be organized consistently. If it is, AI can build a meaningful picture of a book even without deep richness in every record. If it isn't, the AI is working against you.
 

Where the AI Lives Changes What It Can Do

Once you understand that data quality is the real constraint, the question of where your AI lives gets clearer.

When AI is built into the CRM, the output can close the loop. A note gets logged, a record gets updated, a follow-up gets created — directly in the system where the data issue was identified. The data gets better over time. When AI sits outside the CRM, the output has to come back manually. That extra step is where things fall through, and the underlying data problem stays unsolved.

There's also the completeness question. Not all of your CRM data is available through an outside API. By going external, you lose context — the history, the notes, the workflow records that would make the AI output specific to your client. And you're adding a new risk surface: a different vendor, different data handling terms, different questions about what's stored and how. That's not insurmountable, but it's a risk most advisors aren't tracking explicitly.

Trust Is Something You Build Into the System

One thing I worry about is how fast advisors are moving on trust. AI can still make mistakes. It can make assumptions based on incomplete context — and if your data has the kind of inconsistencies we've been describing, you're giving it incomplete context from the start. Without the right evaluations in place before you act on output, you're building on a foundation that hasn't been tested.

Redtail Assistant is a great example of this. When we built it, we had stakeholders from legal, compliance, and security in the room from the beginning rather than at the end. That slowed the build process down more than I'd have preferred, but the guardrails ultimately helped us deliver a better product faster than if we'd had to go back to production after those stakeholders reviewed a complete product. The biggest concrete decision we faced was around personally identifiable information — we shut that off entirely in certain parts of the workflow because, with all of those stakeholders involved, we were able to assess that the risk wasn't worth the benefit at that stage.

We're taking the same approach with Redtail's upcoming AI notetaker — building in guardrails from the start, designed to support firms at different stages of their AI journey. As we talked about this feature with firms, including how far along many of them already are, we could see that the guardrails still matter — they're in place for the firms that need them. But more advisors are open to letting AI reshape how they work than we expected, and that means more firms will be ready to take full advantage of what we're building.

Questions Worth Asking Before You Commit

For any advisor evaluating AI tools, the first question is still data security. Where is your data going? Who's handling it? What are the terms?

The second question, which most people don't ask directly enough: does this actually save me time, and by how much? The biggest value we've seen from CRM-embedded AI is efficiency — meeting prep, note-taking, client data access. That time savings is real. But don't buy the shiny object. Make sure the tool works for your practice and think about tech spread. Nobody wants another login to remember.

The Differentiator Isn't the AI

If everyone has access to AI — and that's where this is heading — AI itself isn't the differentiator. What differentiates is whether your data is in good enough shape for AI to actually use it, and what you do with the time it creates.

The advisors who will be best positioned aren't just thinking about which AI tools to add. They're thinking about their CRM data as the foundation for future conversations — cleaning it up, organizing it consistently, capturing the right context in their notes. Technology gives you a surface. The data is what makes the surface useful.