"By the time I build my practice, will AI have made my expertise worthless?"

An ex-banker asked me that last week, three months from handing in his notice. Twenty-two years behind him, and he is watching people half his age spin up decks, outreach, and content in an afternoon. The question is keeping him awake.

It is not paranoia. The tools are moving faster than anyone can keep up with. Agents are starting to act on their own, reliability is still an open question, and nobody, including the labs building them, can fully explain why these systems do what they do. Six months ago you could hand someone a chat assistant and the risk was low. Now they act.

So if you are weeks away from leaving a 20-year career and betting on yourself, you need the honest answer to his question. It is the opposite of what the fear tells you.

It has made your expertise worth more.

When you ask a capable model to do real work, you get a confident answer in seconds. The hard part is knowing whether that answer is any good. One AI trainer I was listening to this week described being stuck in a loop with a model on an important brief, getting draft after draft that sounded fine and solved nothing, until he stopped, reframed the question, and moved to a different tool. What rescued him was not the model. It was two decades of knowing what “right” feels like, and the judgment to see that the polished text in front of him was wrong.

That is the skill getting more valuable, not less. Taste. Judgment. Knowing what to ask, and knowing whether what comes back is worth keeping. A 28-year-old with the same subscription cannot fake three decades of pattern recognition in a regulated, relationship-driven field. You can.

There is a test from that same conversation that stuck with me. If you use AI to produce something and you cannot explain it without reading your notes back, you should not have used AI for it. The output is only as trustworthy as your ability to defend it in a room. In banking, in wealth, in family office work, that room is the whole business.

Which brings me to the part that matters most for you.

Your clients do not buy your slides. They buy your read on a situation, your discretion, the fact that you have sat across from people making the hardest financial decision of their lives and known what to say. No agent replicates that. In relationship-driven fields, an AI “twin” that speaks to clients for you takes enormous effort to build and still needs a human watching every exchange. There is no version where you buy an agent, point it at your clients, and walk away. That should reassure you and warn you at once.

So here is what I would do if I were sitting where you are.

Pick one AI system and go deep. The advantage goes to the person who has pushed a single tool to its limits. Sampling five of them keeps you shallow on each.

Feed it your judgment, not only your tasks. Give it your best work as examples, tell it how you think, and put it on the parts that drain you: the positioning, the outreach, the first pass at a proposal. I build these as AI agents for my clients so the practice runs without them doing everything by hand. The agent does the lifting. You bring the expertise it cannot have.

And keep the human where the human belongs. The high-stakes call, the sensitive conversation, the final decision. That stays yours.

The advice worth carrying is this: be the best at AI inside your own discipline. You already own the field. You only need the other half.

The wave everyone is afraid of is the same wave that makes deep expertise scarce and valuable again. The people who win the next few years are the ones who already have the judgment and add the tools. That is you, if you move now.

Salama

PS: If you are planning your exit from corporate in the next 3 to 6 months and want to see how your specific expertise becomes an independent practice, I keep a few 15-minute strategy sessions open each week. No pitch, a clear first step. >> Apply here.