I’ve been building software for financial advisory practices for close to 20 years. We use AI in our processes every day now, so I’m not coming at this as a sceptic. I think the technology is remarkable and I’m excited about where it’s heading. But I’ve also watched enough hype cycles in this industry to know that the conversation around AI in advice businesses has drifted into territory that needs some honest pushback.
I like to think of AI as a force multiplier. If the things you’re doing in your practice are already working well, carefully layering AI into those processes will probably make them much better. But if what you’re doing isn’t working – if your processes are messy, your data is fragmented and your team isn’t aligned, AI is likely going to make all that much worse, much faster. That distinction matters more than most people realise.
Where AI helps right now is the boring operational stuff. I don’t mean that dismissively – this is work that eats enormous amounts of time across advice businesses and nobody gets credit for it. Drafting client correspondence, compiling review packs ahead of meetings, sorting through compliance paperwork and following up on outstanding requirements.
We’ve seen practices recover almost a full working day per week by pointing AI at this type of admin. That’s time going back into client work, where it should have been all along.
Meeting transcription is another area where it’s a complete no-brainer. Having client meetings recorded, transcribed and stored as part of your client context – AI is excellent at this, and most firms seem to have gotten onto it already.
Where I get uncomfortable is the growing enthusiasm for pushing AI further into the advice process itself. I’ve said this in a few forums now: AI advice should never become a term. There should always be a human in the loop making the final decision before anything gets recommended to a client. It’s genuinely risky to use software where the AI has full agentic power to run parts of your practice without human oversight.
The reason is fundamental to how the technology works. AI is generative – it generates outputs based on patterns. That’s the root cause of hallucinations, and I don’t think there will ever be a version of the current technology that fully resolves it. So, when you’re dealing with a client’s financial future with suitability assessments and advice that carries regulatory weight under the FSCA, you need a person checking that work. Every time.
I see the same pattern with vibe coding. AI is brilliant at spinning up a prototype: you prompt it, it builds something that kind of works and it feels like magic. But the moment that thing must carry real client data, do actual calculations or keep sensitive information safe, it’s a completely different story. AI is good at creating new things but not great at maintaining them or keeping them secure. If your business relies on it, you need to know what’s under the hood.
There’s also the fatigue factor that I think a lot of us underestimate. We all feel it already. You call a business and you can’t get through the bot wall, it tries to be helpful but sometimes you just want to talk to a person. Advisory businesses are relationship businesses. My own financial advisor runs a structured operation in the background, but he makes me feel like I’m his only client.
There’s never a ticket number, never a “you need to speak to that department”. That’s what good advice feels like, and you don’t get there by automating the human parts. Replace the boilerplate with AI. But don’t try to replace the relationship stuff.
My take is still evolving – it’s all new and it changes every day. But the core of it stays the same. Use AI where it saves your team from work that doesn’t require their professional skill. Keep them on the work that does. And be honest about which is which.
