AI takes over the preparation, not the conversation

At some point I built myself some tools. Not because I am a developer, but because I was tired of losing the same half hour before every meeting.

Look up the company. Skim the annual report. Who will be in the room, what do they do, what have they said publicly of late. Is there a tender, is there a news item, is there an occasion. That is not demanding work. It is simply work you cannot skip without sitting there unprepared.

That half hour is gone today. And for me that is the actual return.

Where the debate usually gets stuck

When I talk to sales people about AI, we almost always end up at the question of whether it replaces the salesperson.

I find the question unproductive. Not because the answer would be reassuring, but because it misses everyday practice.

Gartner describes this under the term AI partnership: a model in which technology does not appear as a replacement but as a partner in the sales process. Classic CRM systems capture and structure data. AI-supported applications analyse customer interactions, prioritise enquiries and prepare conversations.

That is not a statement about the future. It is a description of what has been possible for two years.

What it really takes over

If I summarise what can actually be handed over in practice, it is three kinds of work.

Research. Everything that is public anyway and merely has to be gathered. That is the obvious part and the one with the fastest return.

Summarising. Call notes, minutes, keeping track of what was discussed. That is the work most reliably left undone, because nobody enjoys it and because it is only missed when it is too late.

Prioritising. Which enquiry first, which existing customer is showing movement, where is something taking shape. The bigger lever sits here, because AI sees patterns that escape a person in the daily run. And the bigger risk sits here too, because a prioritisation nobody questions turns very quickly into a self-fulfilling prophecy.

And what it does not take over

Interpretation.

An example from everyday work. Before a meeting the AI gives me the information that the company is currently building a plant in Eastern Europe. Correct, found, cleanly summarised.

What it does not give me is whether I should raise it. Whether the person opposite is proud of it or whether it is political. Whether the build is carrying their career or costing them sleep. Whether they are even allowed to talk about it yet.

That judgement forms in the first two minutes of a conversation, out of tone, hesitation and what somebody does not say. Gartner calls this part mentalizing, the ability to read and assess the behaviour and feelings of the person opposite. It is the part that does not become less important through automating the rest, but more.

That is the point I consider central: the more preparation is automated, the more the close is decided by the part that cannot be automated.

The uncomfortable condition

For this to work, you have to do something many organisations have no budget line for. You have to teach it to people.

Not the operation. The operation is trivial. Two other things.

First, the ability to ask precisely. What the debate calls prompting is at its core an old sales virtue: framing a question so that the answer is usable. Whoever can do that in a customer conversation can do it here.

Second, and more important, the ability to distrust results. An AI-generated call guide always looks good. It is fluently written, well structured and sometimes simply wrong. Anyone carrying it unchecked into a meeting appears confident and is wide of the mark, and that is a worse combination than unprepared and honest.

A good part of the training therefore belongs not in the application but in the questioning of it.

What I took away from this

AI has not taken a single decision off my hands in sales. It has given me time back, at exactly the point where I needed it least.

So the interesting question is not what the technology can do. It is what you do with the time you gain. If it flows into more meetings, you have an efficiency gain and nothing else. If it flows into better meetings, you have changed something.

I built the tools myself, which is why I know fairly precisely where the benefit sits and where the limit runs. The limit runs where somebody has to decide whether to say a sentence out loud or not.

What do you use the time you gain for?