only if you give the technology some real context— a real situation with an actual goal. The answer gets dramatically better.
What to do: Try writing a prompt for your AI, then ask it what you can do to make the prompt even better. It will rewrite it for you in its own language!
4. Bots don’ t know what happened this morning.
Most AI models are frozen at a certain point in time. If you ask them about yesterday’ s Fed announcement or this week’ s client is scared. Doesn’ t know she just buried her husband three weeks ago and is trying to hold her emotions together on a phone call about beneficiary forms.
AI also has no idea what happened in the five minutes before you opened your laptop. No idea about the shake in a client’ s voice. No idea that a client’ s silence on a call means something. It cannot read a room it was never in.
Yet it will write you a technically correct, perfectly polished paragraph either way. Completely blind to the feelings of
Before you copy anything into an AI portal, ask yourself: Did my firm choose this tool? Perhaps because it’ s a version that protects client data? Or is this a free app I downloaded because it was easy?
8. Agentic AI can take action without asking you first.
There is a real difference between a chatbot that answers a question and an agent that does something( updates your CRM, sends an email, checks a calendar) without you approving every step. That’ s what agentic AI does, and it’ s a powerful tool. And yet even if you put safeguards in place to keep this technology from doing things you don’ t want it to, the agentic bots will make mistakes, just like humans do.
That’ s a different category of risk than typing a prompt and reading a response.
What to do: Before you turn any agentic tool loose, know exactly what it is allowed to do on your behalf. Find that out before it acts, not after. Or even better, don’ t build one unless you have gotten some real education on how it works. market move, some tools will just guess. Or worse, they’ ll answer confidently with old information. Surprisingly old.
What to do: If you’ re looking for anything time-sensitive— current interest rates, breaking news, recent filings— use a tool that actually searches live. Never trust a static model for something that changed an hour ago.
5. AI forgets what happens the second you close the tab.
They say certain animals, like bees and snakes, have very short memories. AI can be like that. If you start a new conversation with a bot tomorrow, most of them will have zero memory of what you talked about today. You are starting from scratch every single time, unless you build something that carries over.
What to do: Save your standard prompts, your repeat tasks— save all of it as a project in Claude or ChatGPT( if they’ re approved for your firm). When your work is saved as a project, all the context remains. Stop reintroducing yourself to the tool every morning.
6. AI has no idea what just happened in the room.
This technology doesn’ t feel anything. Not a little. Not at all. It doesn’ t know your the human being reading it, perhaps that person who was recently widowed.
That is exactly why this is your job to write letters to such clients. The AI might be able to draft words. But it’ s you who will decide which words actually belong in front of this person, today, in this moment, given everything you know that the AI never will.
What to do: Obviously, you should never send an AI draft letter straight to a client without reading it first and trying to hear it the way they would. If you wouldn’ t say those exact words out loud to that person’ s face, then you shouldn’ t let the AI do it on your behalf.
7. AI can’ t design your workflow for you.
Let’ s say you download one AI tool and then another. What if none of them do what you thought they would, and you walk away thinking AI is not for you?
The problem might not actually be the AI. It might be that you’ re asking it to organize your time in a way it can’ t( and wasn’ t built for).
What to do: Pick one workflow— meeting notes or follow-up emails— and stick with just that one for 30 days. Get good at it before you merely start adding tools to solve problems that AI doesn’ t solve.
9. Yes, you can use AI and still be compliant with rules and regulations.
I know nobody wants to read the compliance section. But stay with me.
There’ s a myth that advisors are the ones jumping ahead into the technology and the compliance teams( and home offices) are playing catch-up trying to keep advisors in line.
What I’ ve seen in my visits with firms is usually the opposite problem: Most advisors are scared to even bring AI up with their compliance teams, assuming that if they ask to use it, the answer will be no. So they either ignore it and fall behind or they quietly use it on their own off to the side without telling anyone. Neither approach ends well.
Those who avoid AI entirely lose time and a real advantage while competition down the street gets faster and sharper every week. The ones using it quietly are worse off. Some advisors I’ ve met have built an entire workflow around artificial intelligence, using it for meeting notes, follow-ups, prep work— the whole thing— only looping in the compliance team three months later when someone finally asks an uncomfortable question. By then, it’ s not a conversation. It’ s a cleanup job. The firm’ s compliance must then unwind something that has already touched
SEPTEMBER / OCTOBER 2026 | FINANCIAL ADVISOR MAGAZINE | 31