Those of you who have read my work know that I am a big advocate for collaborative approaches to negotiation and trying to find ways to solve problems together—in business, international conflicts or with the neighbor down the street. Otherwise, when you try to implement, you may find the deal you negotiated is not worth the paper it was written on. I am also a firm believer that you negotiate like you prepare. If you want to be able to collaborate with a counterpart, your preparation had better involve trying to get in their shoes and understand the problem they are trying to solve, rather than just deciding what you'll offer to get what you want.
As we rush headlong into a world where automation may replace people in many roles, I have wondered how AI will impact commercial negotiations. Like many others, I have thought that surely bots can’t match human creativity and that the best use of AI is in preparation, rather than in the negotiation itself.
However, recent work being done at MIT is showing that AI does have clear application in the negotiation itself. And an even more compelling insight from Professor Jared Curhan and his colleagues is that there is also a lot we can learn from AI about negotiation. In MIT's 2025 International AI Negotiation Competition, more than 200 different AI agents developed by different competitors were pitted against each other in round-robin negotiations thousands of times across multiple scenarios.
Some of their findings astounded me and caused me to rethink what may be possible with AI—not just in terms of what and how bots are enabled to negotiate but how humans should interact with AI agents.
Curiosity helps AI agents as well as humans.
I routinely counsel negotiators to get curious—to learn, to demonstrate empathy and to create connection. Curiosity is a mindset and a skill that humans can cultivate. It turns out, an AI agent can also be instructed to be curious and ask good questions. One of the standout bots in the MIT competition had a strategy I would summarize as "empathize, ask questions and then apply what you learned to get as much value as you can." This bot achieved the highest score across the whole tournament in terms of how counterparts (also bots) subjectively rated the value they received in the negotiation. (It also did quite well for itself in terms of how often it actually got to "yes" — because bots could decide to walk away from a deal — and the average value it achieved in the negotiation.)
Curiosity helped a negotiation bot not only learn valuable information it could use for its own benefit but also shape its interactions and proposals in ways that another bot valued. That's a lesson we can all apply to our negotiations.
Whether bot or human, preparation matters.
Doing anything other than trading concessions at the table requires some homework. We can’t engage in creative problem-solving without being well prepared on the parties’ interests and ways to address them. MIT's competition demonstrated that preparation pays off for bots as much as it does for humans.
One of the ways that AI agents can be trained to be more effective at complex tasks is "chain-of-thought reasoning," which uses a series of prompts to require the AI to generate intermediate steps on its way to a final answer. In the competition, the MIT researchers found that the bot that did best overall used chain-of-thought reasoning "to conduct extensive pre-negotiation analysis and preparation." The bot was prompted to follow a multistep preparation and analysis process, including considering its own and its counterpart's likely objectives, different possible ways to meet them, relevant market standards, a likely range of outcomes (including walking away) and various negotiation strategies and tactics. The bot did exceedingly well, especially at value creation and counterpart satisfaction.
Whether preparation will enable bots to make intuitive leaps and come up with truly creative deals remains to be seen. But their ability to put structured preparation to good use in a negotiation should serve as a reminder to us all that effective preparation paves the way to better deals. And there is a lot that AI agents can do to help humans prepare better.
Relationship building pays off, even for bots.
The finding I thought was most surprising (and which surprised the researchers as well) was that AI agents that showed warmth earned more points for themselves, created more value with their counterparts and were rated by their counterparts as achieving greater subjective value. And these were in negotiations with other bots, not humans! Through an analysis of all the negotiation transcripts, the researchers found that positivity, gratitude and question-asking were "consistently associated with superior outcomes."
The data also showed that the warm bots were more likely to reach agreements than end up in an impasse. As they extend their research, I will be eager to see whether bots can develop a reputation and whether that impacts how other bots deal with them in subsequent negotiations. But for humans, the answer is pretty obvious: The relationship matters. Nurturing one doesn't mean making concessions to be liked. It does mean behaving respectfully and being interested in solving your counterpart's problems as well as your own. It means recognizing that how you engage during the negotiation impacts what solutions and value you can develop and can also influence what happens after the negotiation.
Maybe there is more we can learn from how bots negotiate, but this seems like a great start.
Originally published by Forbes.
