Dear Negotiation Explorer,
I built four negotiators this week, and not one of them is a real person.
I built them because before I sit down with a new client I want to understand how their work actually runs, and there is no way to learn that from a website, so I made people I could interview.
Two of them are a seller and the buyer sitting across from her on the same deal, built out of research on a real company I am working with. Sofia Lindberg sells industrial cooling plants. Matthias Keller buys them for a cold storage operator.
The other two are the two sides of a case I have been teaching for a few months. Sunita Deshpande, chief procurement officer at PharmaCrest, nineteen days away from a contract worth eighteen and a half million dollars renewing itself automatically. David Morrison, who has to keep it.
I interviewed all four, one at a time. No session saw any of the others.
Nine problems came up. Every one of them showed up on both sides, buying and selling.
What I built
The agent is called Understudy. In a theatre, the understudy is the person who has learned somebody else's part well enough to walk on and play it.
You give it what you already know about a role. The market, the pressures, how the work runs, a meeting note if you have one. It writes a character out of that material, and then you sit down and talk to them.
Every trait has to trace back to a line in a source, so a character cannot want something merely because it would be interesting. And a program checks the character before you can use it. Mine rejected the first two versions of one of these four, and I did not overrule it.
This is not the counterparty simulator I already run. Von Neumann plays the person across the table in a deal you are already preparing for, and it needs a real, named counterpart and a finished preparation report before it can start. Understudy gives you somebody to interview when you have none of that yet.
Three problems in how they prepared
The hardest negotiation was inside their own company, and none of them called it a negotiation. Before Sofia can put one package in front of a customer she has to collect four separate agreements from colleagues in three departments who do not work for her. She had always thought of that as coordination. Matthias has the same problem from the buying side: the savings number belongs to finance, the technical judgement to engineering, the date to production, and the award file to him.
The deadline came from somewhere else, and it took away their room to move before anyone spoke. Matthias inherited a go-live date that production had set before the requirement reached him. He told me his best lever in any negotiation is being able to wait, and the calendar removed that before he said a word about price. Sunita has nineteen days on a clause her team found three weeks ago and the other side has known about since signing. Nineteen days is not enough to stand up an alternative.
None of them could build a scenario of the other party. Not one of the four had written down what the other side wants, ranked in order of importance, or what that side would do if there were no deal. Sunita gave me four scenarios and every one of them was about how she would react to David. Sofia never asked what her customer's real driver was; she assumed price, because price was what he showed her. Matthias never once considered what his supplier would do without the order. A hypothesis about somebody else is not a fact, and in their world anything that is not a fact cannot go in a file, so it never gets written.
Three things they did not know about their own company
The cheap choice needs no explanation and the better one needs a case. Matthias is measured on one thing, the saving he books. Nothing else he cares about is written down anywhere, so the cheapest offer is the only one that meets a target somebody set him. He had a cheaper offer that met the specification on paper and did not want to buy it, and he could not write down a reason that would still hold up if somebody opened the file four years later. Awarding anything else requires an argument, written by somebody qualified to write it. Sofia has that argument. It lives with her technical and service colleagues, and it reaches her after the meeting where she needed it. The only objectives any of the four had in writing were numbers.
Nothing they learned survived the deal. Sofia cannot tell you why she lost her last three deals, because nothing anywhere records the reason. Every deal produces a competitor comparison, a list of objections, and a set of terms that were accepted or pushed back. When it closes, all of that stops existing. Matthias does not know whether his own past awards can be searched at all.
None of them could say what they were allowed to agree to. Sofia knows there is a price below which she can move on her own, and she cannot tell you where it is. Matthias knows there is a figure above which the award is not his to sign, and he could not recite it either. Sunita cannot say how much of the fee her finance chief will let her put at risk. David is carrying two different numbers, what he is authorised to agree and what he wants to land.
Three things they had wrong about negotiation itself
Everything collapsed onto the one issue with a number on it. Sofia lost a deal the day a buyer pushed a cheaper quotation across the desk and asked her to explain the difference. Her argument about running cost, electricity, and how fast somebody arrives when the plant stops on a Saturday was real, and she had no figures in her hand, so it sounded like a salesperson talking. Matthias described the same moment from the other chair: a real number against a set of ifs, and the number wins by default unless somebody does the work to make the ifs concrete.
Research shows this is not people deciding that money matters most. In large experiments where the same trade-off is written out twice and only the side carrying the figure is swapped, people shift towards whichever side has the number, while telling you both things matter equally. Numbers are easy to compare, and sentences are not.
All four collapsed the other side's alternatives into tactics. An alternative is a BATNA. It is what the other party does if there is no deal at all, away from the meeting, whether or not the two of you ever speak. Sunita listed the moves David might make in the room and called those his alternatives. What she had never asked is what he does if she gives notice, which is the only question that tells her how badly he needs her. Sofia has the same gap from the selling side: she prepares against a competitor's headline price, because a headline price is all she ever sees.
The creative option existed, and it died before it became a proposal. Sunita circled a joint venture with McKinsey for twenty minutes without landing it, then attacked it the moment it was named, before she had designed it. David had the same idea sitting inside his own answer and was using it to defend himself rather than building his offer around it. Sofia invents good trades live in the room and cannot arrive with one already agreed, because agreeing one in advance means asking three colleagues to commit to something that has not happened yet.
What a scenario is actually for
Sunita got there before I did. After I sent her scenarios back a second time, she stopped defending them and said they had a second job. They were questions she would go in and test.
Under each one she wrote down what would prove her wrong. If he raises the notice date himself in the first ten minutes, she is in one world. If he asks what else she has been looking at, she is in another, because somebody sure of his position has no need to ask.
She could not do that an hour earlier.
Understudy cannot tell you what a real buyer wants. Every one of the four gave me a list of things only a real person can answer.
What to do before your next meeting
Before your next real meeting, write down what you think the other side wants, in order. Write what they would do if there were no deal with you. Then write the one question that would tell you whether you have it wrong. You do not need an agent to do it.
This week's question
What is the one thing you are assuming about your counterpart in your next negotiation, that you have never actually checked?
Just hit reply and tell me. I read and answer every one.
#Negotiation #NegotiationSkills #NegoAI #AI #AIAgents #AugmentedNegotiator