Dear Negotiation Explorer,
Seven weeks ago I told you I was building the NegoAI platform in the open, every week, whether it went well or not.
I have my own server in Germany, €12.99 a month. An agent called Alex lives on it. The chat where I talk to her runs on the same machine, so my deal information never goes near anybody else's company. Alex has a memory now, and next to it a file that holds only the things we do not know. That second file is the one that turned out to matter.
She has her own email address, her own face, her own voice. She reads my inbox and my calendar and she cannot send anything.
Then there are four negotiators who are not real people. We built them so I would have somebody to interview before I sit down with a client whose business I do not understand yet.
One page gets me ready for that first meeting.
And there is a second team of agents whose only job is to build the platform. They have now run a whole job on their own, with me in none of it.
This week I went back to those four conversations.
Nobody had thought about the other side
Sunita buys consulting for a pharmaceutical company. Eighteen and a half million dollars renews itself automatically unless she gives notice. Her team found the clause three weeks ago, so she has nineteen days. David is the partner at the consulting firm. His job is to keep that contract.
I asked her for her scenarios and she had four ready. He plays the clock and lets the nineteen days run out. He drops the price and changes nothing else. He offers her a better structure in exchange for a longer term. He has decided she cannot really leave.
Good scenarios. Every one of them is a plan for what she does back.
Sofia sells industrial cooling plants. A buyer turned a cheaper quotation round on the table and asked her to explain the difference. Everything she wanted to say about running cost and response times had no number on it. She lost that deal. She will not tell you the two things are connected, because she cannot prove it.
"I assumed price because they showed me price."
Matthias buys the same equipment for a cold storage operator. His best lever in any negotiation, he told me, is being able to walk away or wait. Production set his go-live date before the requirement ever reached him, so the calendar took that lever before he said a word about price. What his supplier would do without the order never came up once.
And David had been preparing for two weeks. The first thing I asked him was what Sunita wants.
"I've been building my own case, not hers."
So the bigger deal died in the same place every time
Sunita circled a joint venture with the consulting firm for twenty minutes. Then somebody named it and she attacked it, before she had designed any of it. David had the same idea inside his own answer. He was using it to defend himself instead of building his offer around it. Sofia invents good trades live in the room and can never arrive with one already agreed. Agreeing one in advance means getting three colleagues to commit to something that has not happened yet.
Every one of them reached the bigger deal and then killed it.
You cannot build a trade until you know what is cheap for them to give and expensive for you to get. None of them knew, because none of them had looked.
Then I asked them, and all four could do it
I told David to sit in her chair. He came back inside the same conversation with three different Sunitas. One is protecting the system her own people built. One needs a number she can defend to a board. One is simply angry about twelve years of being under-served. Each one came with her interests in order, what she does if there is no deal, and the tell that says which one he is sitting with.
Twenty-nine years in that firm, two weeks of preparation, and none of it had been spent on her side.
I sent Sunita's four scenarios back and told her they have to be built on David's interests and David's alternatives, not on her own reactions.
"Earlier I built packages around my levers. Now I've built scenarios around my reactions. Both are me looking at my own side of the table with better lighting."
She rebuilt all four from his side, then went further than I had. The scenarios are not predictions, she said. They are questions she goes in to test, and under each one she wrote the thing that would prove her wrong. He raises the notice date himself, unprompted, in the first ten minutes. He asks what else she has been looking at, which a man certain of his position has no need to know.
An hour earlier she could not have done any of that. Nothing had changed except that somebody made her.
What surprised me
They are the negotiators I sit with every week.
I am in these rooms with people who sell and people who buy, in companies much bigger than the ones I invented. Same thing every time. Nobody has the other side written down, so the scenarios are all about their own moves. The bigger deal gets named and then killed. And every one of them can do the work immediately, the moment you sit them down and make them.
Nobody put that in. Brama, my partner in Claude Code, did the building. We gave each of them a market, the pressures they work under, how the job runs, and a meeting note. Nothing about how to prepare. It came out on its own, four times, on both sides of two different deals.
Four is four. I am not going to pretend that is a study.
But it did not come from them. It came from the material they were made out of. That is how this job gets written down and passed on. Nobody teaches this part. Where would you have learned it?
They told me what a real person never would
Sofia told me she does not know where her own authority on price stops. She has a feeling for it, she said, not a rule. Matthias told me there is a figure above which the award is not his to sign. He has never seen it written anywhere he could repeat back to me.
Nobody says any of that in a real meeting. Not to me, and I think not to themselves either. Part of it is pride. Part of it is not wanting to say something in front of somebody who might repeat it upstairs.
Which means the gaps are the same on both sides of your table and neither of you will show them. That is the whole reason preparation cannot be a set of facts you go and look up. It is what Sunita ended up with: a hypothesis you walk in and test.
What we built against it
Three things answer those three gaps: one I already had, and two are in the list at the top. I did not plan that. Von Neumann plays your counterpart in a deal you are already preparing for, so you test their side instead of imagining it. The page settles the questions before the meeting instead of during it. And one of the builders exists only to argue against us before anything goes out.
Somebody has to ask you, so pick somebody
Three things this week, before your next real conversation. None of it needs software.
Write down what the other side wants, in order. Not what they told you. What you think, ranked, so you can be wrong about something specific instead of wrong about everything.
Write down what they do if there is no deal with you at all. Not their moves in the room. What happens to them on Monday if you both walk away.
Then give both lists to a colleague and have them ask you how you know. That question is the only thing that changed for these four, and it costs nothing.
That's it.
What's next on the calendar
The autumn lightning lesson series opens in September. Four sessions, and one registration covers all four: 10 and 24 September, 8 and 22 October, Thursdays at 18:30 Rome time.
This week's question
Who has ever made you write down what the other side actually wants?
Just hit reply and tell me. I read and answer every one.
#Negotiation #NegotiationSkills #NegoAI #AI #AIAgents #AugmentedNegotiator
