← All essays

What happens when your buyer knows more about you than you know about them?

Your next proposal may never be read by a human.

An agent receives it. It compares it against four others, checks your references, reads your terms and rules you out. Or it does not. The whole thing takes less than a minute. No meeting where you can read the room. No call where you can set something straight. No person who remembers the last ten years.

That sounds like the day after tomorrow. It is already happening, and it already has a name.

You are entering a race that is already over

There is a moment everyone knows who has been in sales long enough. The conversation goes well, the questions are friendly, everyone nods. And still you sense something is off. As if you were playing a part in a play whose ending is already written.

That feeling is not deceiving you.

By the time your buyer first calls, most of the decision sits behind them. Their shortlist exists, ordered by preference, and it is older than your meeting. In 95 out of 100 cases they end up buying from a vendor who was on that very first list.

That is the one number worth remembering. Everything else follows from it. The meeting you were invited to is no longer an opportunity. It is a confirmation.

The advantage that no longer exists

For decades this profession rested on a quiet agreement: the seller knows more. About the product, the market, the alternatives. You met because there was no other way to get at that knowledge.

That agreement has been terminated. Not by your customers. By a machine.

One in two B2B software buyers now starts research with an AI chatbot rather than Google. A year ago it was one in three. And seven in ten ended up choosing a different vendor than planned, because the assistant took them there.

Let that last sentence stand for a moment. Seven in ten decisions were redirected in a conversation nobody from your side attended. There was no meeting you could have prepared better for. No objection you could have countered.

A seller who delivers information is a slower search engine.

And now the machine buys

So far this has been about a human preparing with AI. That is the harmless version.

The more unsettling question: what happens when the machine does not just research, but decides and orders?

This is not speculation. The survey measuring agentic AI in purchasing lists as the most common use cases, verbatim: searching, evaluating and choosing products. Viewing order statuses. And then, without any fuss at the end of the same list: ordering products and services.

An agent cannot be persuaded. It has no good days and no bad ones. It does not remember that you delivered over Christmas three years ago when things got tight. It compares what is comparable, and everything else does not exist for it.

That is where your future is decided. If you have nothing to offer beyond what fits in a table, you will be compared in a table. And in tables, the cheapest wins.

Off-the-shelf AI gets you into the game, not ahead

The obvious reflex: so we buy AI as well. Do that. Anyone not using AI in sales today is losing pace, immediately. It is the right first step, and there is no way around it.

It just remains a step everyone takes.

I have seen AI from the inside, during my time as Chief Customer Officer at Aleph Alpha. Not as a user rolling out a tool, but while building it: I sat in the rooms where it was decided what a model must be able to do and what it never will. What I learned there is uncomfortable for anyone hoping the tool will do it. Both sides buy the same technology, trained on the same internet, with the same answers to the same questions.

A Gartner figure shows how little is left. Buyers were asked who they expect misleading information from. 51 percent said generative AI. 49 percent said the sales rep. In other words: your best seller is barely more trusted than a machine.

Off the shelf fits a lot of things. Never the decisive one. A tool everyone has sets no one apart, and a model anyone can rent is not an advantage, it is an entry ticket.

AI gets interesting the moment you connect it to what only you have. Your market. Your customers. Your history. The reasons decisions fall the way they do in your business. It is not the model that makes the difference, it is the knowledge you put into it.

What nobody can take from you

Think of the person in your sales team who has been there twenty years. Who reads an enquiry and says after thirty seconds: we are not bidding on that. Ask why and you get a shrug. They just know.

They do not just know. They earned it across hundreds of proposals, won and lost. It simply never occurred to anyone to ask them about it.

Why you won the deal in March and lost the one in May. Which objection comes up every single time, and at minute forty rather than minute ten. Where your price ceiling actually sits.

That is what I call the unwritten. No model has read it. No competitor can buy it, because nobody sells it. And it carries an expiry date that appears in no calendar: the day that person retires.

Here the circle closes back to the agent. The more the other side decides automatically, the less your presentation skills count. The more it counts what you know about your own market and build into your offer before anyone compares it. The unwritten does not become less important when machines buy. It becomes the only thing that counts.

The uncomfortable punchline

And now the part almost everyone skips.

Your AI knows the internet. It does not know your business.

It does not know why the March deal came and the May deal went. It does not know your competitor, your price ceiling, that one objection. To your specific questions it gives fast, general answers. That feels like progress, which is exactly what makes it dangerous.

In Germany, barely one in ten AI-using companies applies AI to internal knowledge management. The knowledge that decides deals still sits in people's heads and in systems that do not talk to each other.

So the task looks different from how it is usually framed. Not roll out AI. Capture your own knowledge in a form a machine can work with.

That is uncomfortable. It cannot be finished in a quarter and cannot be bought as a licence. A tool is rolled out by Friday. For the unwritten you need people who write down what they previously only knew. And a reason for them to do it.

There is still time. Not much.

One closing observation you will like.

In the US, buyers already use agentic AI more often than suppliers do. In Germany it is still the other way round: sales uses AI more often than procurement.

So your procurement is not ahead of you yet. Anyone reading that as reassurance looked at the wrong number. The relevant one is the other: research starting in a chatbot nearly doubled in eleven months. A movement at that speed does not stop at a border.

You do not have a problem. You have a window.

What you do in that time does not decide whether you have modern tools. Everyone will have those. It decides whether, in three years, you still offer something your competitor cannot order from the same vendor.

Everything else your buyer has already read. Or their machine has.

In short

Why does the buyer know more than the seller today?
Because they make most of the decision before making contact, increasingly with AI. At first conversation they are on average 61 percent through their buying journey, and 94 percent of buying groups have already ordered their shortlist by preference. The seller walks into a decision that has largely been made.
What changes when AI agents do not just research but order?
Everything based on persuasion loses value: presentation, relationship, reading the room. An agent compares what is comparable. Deloitte Digital already lists not only searching and choosing products as use cases for agentic AI in purchasing, but ordering as well. If you have nothing to offer beyond comparable specifications, you will be compared on price.
Does using the same AI as the buyer help?
Using AI is the right first step, but it creates no advantage, because both sides buy the same technology trained on the same public data. Gartner finds buyers consider generative AI (51 percent) and sales reps (49 percent) almost equally likely to mislead. The difference only appears once the AI is connected to your own market, customer and company knowledge.
What is the unwritten?
The knowledge about your own sales motion that appears nowhere on the internet and therefore sits in no general AI model: why one deal was won and another lost, which objection comes up every time, where the price ceiling actually sits. It is the only advantage left when both sides run the same AI, and your own AI knows it no better than the buyer's does at first.

Sources

  1. 6sense: B2B Buyer Experience Report 2025, 12 November 2025

    At first contact, buyers are on average 61% through their journey (2023 and 2024: 69%). 94% had ordered their shortlist by preference before engaging any vendor; for 77% the first vendor conversation was with the eventual winner; 95% of the time the chosen vendor comes from the day-one shortlist. Nearly 4,000 buyers: 46% North America, 20% continental Europe, 20% UK and Ireland, 14% Asia-Pacific.

  2. G2: G2 2026 Buyer Behavior Report, 15 April 2026

    51% of B2B software buyers now start research with an AI chatbot more often than with Google, up from 29% eleven months earlier. 69% chose a different vendor than initially planned based on chatbot guidance. 1,076 B2B decision-makers, fielded March 2026.

  3. Deloitte Digital: Accelerating sales growth through B2B digital commerce, January 2026

    38% of B2B buyers use agentic AI in the purchasing process, most commonly to search, evaluate and choose products. On the supplier side it is 24%. 530 respondents each, US, director level and above, companies with 1,000+ employees, fielded August/September 2025.

  4. Gartner: Gartner B2B Buyer Survey, 20 May 2026

    51% of B2B buyers say they are more likely to encounter misleading information from generative AI, while 49% say the same about a sales rep. 645 B2B buyers, fielded August/September 2025.

    Gartner blocks direct access to the press release. The wording was verified against two independent, verbatim redistributions of the same release. Volltext

  5. Bitkom: Künstliche Intelligenz in Deutschland: Perspektiven aus Bevölkerung und Unternehmen, 15 September 2025

    Only 11% of AI-using companies apply AI to internal knowledge management. Base: 215 AI-using companies out of 604 surveyed by phone with 20 or more employees, fielded weeks 27 to 32 of 2025.

  6. Institut der deutschen Wirtschaft (IW Köln): Künstliche Intelligenz als Wettbewerbsfaktor für die deutsche Wirtschaft (IW-Report Nr. 33), 4 July 2025

    Among AI-using companies in Germany, 45.5% apply AI in production and 29.3% in sales (top-5 table, n=428); procurement does not appear in the top 5 and therefore sits below 26.8%. For planned adoption the gap is stated explicitly: sales 43.6%, procurement 33.0% (n=239). IW-Zukunftspanel wave 49, 1,038 companies, fielded 2024.