HICO Group AG

The easy Source of Wealth file looks like a payslip with better stationery. The one that decides whether AI is useful looks like this: a company sold in one country, dividends booked in another, a trust mentioned in the second meeting, and three fluent versions of the same life in three systems. Compliance asks for the document you already sent. The client only feels the delay. That is the file I would use as the test, not the salaried profile that will look fast in the first demo. An agent can assemble the evidence and draft the biography. It cannot decide whether the origin of the wealth is actually understood. If the banker still cannot explain the conclusion without the prompt history, you have not shortened onboarding. You have written the same unfinished story at a higher word count. Which file is your house using to judge the pilot: the clean one, or the one the client would rather not have written?

My daughter is currently looking for a princess in an unfinished castle. We are at Neuschwanstein. From the outside it looks complete. Inside, stairs stop and rooms were never built. She still expects the princess behind the next door. I did not have the heart to tell her the princess is probably still waiting for the architect. I see the same pattern with AI. A system is announced, demoed, and shown to the board. The facade is impressive. The connections that would let people actually use it are still missing. Context lives in people’s heads. Ownership is unclear. The model answers anyway, with the confidence of someone who has never seen the missing stair. The clear move is to name the gaps before the model is allowed to speak. What was never built, who owns it, and where a person still has to step in. Otherwise the explanation only arrives after someone has already trusted the answer, which is an expensive way to discover that the tower was a sketch.

Companies keep buying AI licences. Then they wonder why half the building hides the work in a private chat. New data keeps pointing at a cause most boards still treat as soft. Empathy. Not as a poster on the wall. As whether people believe the organisation will still want them after the tool works. In empathetic cultures, almost nine in ten employees say they got decent AI training. In toxic ones, a third. Eighty percent in the first group say AI gives them more control over their work. Just over half of all employees say the same. A third of C-suite respondents still name headcount reduction as the main reason they invest. Sixty percent then admit they will need more empathy as the firm depends on the machines. That gap is the strategy. I saw the practical version of this in Singen last week. We can give people agents that plan, build and test. We can put a human at the gate so nothing goes live without a yes. None of that scales if the room thinks the project is a polite way to shrink the team. They will not bring the real process. They will keep the clever version in a notebook. I watched the same pattern in banking. Internally focused houses talked transformation and measured cost. The people who still owned the client relationship heard a different message. Adoption stalled in the hallway, not in the architecture. Culture is not the opposite of governance. It is what makes governance usable. If leaders cannot say which work stays human, who is allowed to build, and how people will be measured after the speed arrives, the model will be used against the grain. Quietly. Fast. The question I would put to a leadership team this week is not “have we rolled out AI.” It is whether a capable person on your team would show you what they already built.

The AI policy went live in March. The good ideas still live in a notebook named “do not share”. That is the follow-up I keep getting after writing about culture and AI in banks. Not a debate about models. A quiet admission that people already built something, and they will not put it on the shared drive. I understand why. For years the next programme meant more controls and fewer seats. Empathy research now says the same thing in survey language. If you experience your leaders as being on your side, AI feels like control over the work. If you do not, it feels like a test you can fail by being visible. So the interesting question on 7 October is not “did we communicate the strategy.” It is whether last week’s best process still has a filename the company is not allowed to see. If the answer is yes, you do not have an adoption problem. You have a contract problem. And no agent will fix that.

Gartner’s useful line on AI in sales is simple: redesign the decision, not only the task. Private bankers were measured on NNA. Our sales managers are measured on revenue. AI only helps either number if it changes which conversations happen this week, not how many emails leave the building. For us that means AI does four jobs on every signal that reaches commercial growth: - Detect the change - Score how fresh and how strong it is - Draft the why-now line - Write the audit trail A person still does the fifth: decide whether we act. HICO Forge is the place those five steps should live. Not as another dashboard. As the operating model behind Awareness, Buying Process and Customer Growth. Agents prepare. They queue. They do not send, and they do not close the trust gap on their own. Buyers still want a human at the moment of confidence. The teams that hit the number will not be the ones with the most agents. They will be the ones that decided which signals an agent is allowed to move, and which decisions stay slower on purpose.

The signal everyone buys is external intent. The signal most sales teams ignore sits in the project. That is also where AI can do work that a purchased intent feed never will. A consultant documents a new pain. A quarterly account meeting surfaces a second use case. NPS drops. A renewal conversation gets tense. Those are first-party signals with context attached. I saw the same in banking. The conversation that moved NNA rarely started with a cold briefing. It started with something that had just changed inside an existing relationship. Selling data and AI services, we see it again. The next revenue is often already sitting in an account we serve. This is the part agents should read first: project notes, meeting minutes, NPS comments, delivery friction. Not only job ads and category surges. AI can surface the sentence that means they are ready. It cannot be the only one who notices it. That still needs a person who was in the room, and a system that does not let the note die in a folder. If that loop is only a slide in a steering meeting, sales managers will keep hunting strangers while the quarter is hiding in the current book.