Why AI Gives Negotiators 48% Better Outcomes (Even Without Trust)

In this first of three episodes based on my conversation with negotiation expert Remi Smolinski, I share key insights about how AI transforms negotiation outcomes. Based on research with 120+ executives, we discuss why negotiators using large language models achieve 40-48% better individual results and 84% better joint outcomes - even without building trust first. I explain the practical workflow systems I've developed for sales and procurement teams and why symmetric AI adoption creates fairer negotiations. Key findings from my paper 'When AI Joins the Table: How Large Language Models Transform Negotiations.' 

If you enjoyed this episode, please leave a review and check out our website: www.negoai.ai

I welcome any suggestions, questions, or comments at yrana@negoai.ai

--------------------------------------------------------------------------------------------------
Copyright Disclaimer Under Section 107 of the Copyright Act 1976, allowance is made for "fair use" for purposes such as criticism, comment, news reporting, teaching, scholarship, and research. Fair use is a use permitted by copyright statute that might otherwise be infringing.

Transcript

[Remi]
Tell us a little bit more about, uh, what it has led you to. Uh, about your project, what's your projects, what you're doing, what you're focusing on, both in terms of, uh, in terms of your advisory work, uh, as well as in terms of your research and teaching. 

[Yadvinder]
Uh, in teaching, uh, this since this year, I've been asking my students to develop a chatbot for cross-culture negotiation. And some of the works and some of the projects have been amazing, uh, amazing. Incredible. So I think it's a good way for them to learn more about, uh, the concepts and the, the theory behind cross-culture negotiation, but also to apply it. Uh, and today, I think, uh, having AI literacy is fundamental. And this is my starting point in, in class. Uh, we're quite free in our university to use AI and, uh, to develop tools with AI. Uh, in my advisory, I'm a... Both with sales and procurement teams, I'm helping them to prepare better for negotiation, uh, in more efficient way, getting also better results. Uh, so I have a workflow of different agents that combined. Uh, for example, for a sales team, provide the value proposition, uh, that is linked to a negotiator, that is linked to a LinkedIn profiler, and that at the end provides a report for the sales rep before a meeting. Uh, and the same, we can do it for procurement. 

[Remi]
That's absolutely amazing and, uh, I've already praised this idea, uh, in, in our previous conversation. I, uh... Let me, let me do it again here in front of our audience, viewers and listeners, uh, because it's, um, it makes a salesperson or, or a procurement person much smarter. But you also, you're also a researcher, Edwinder, yes? So I, um, I read your paper, uh, recently, the one you published on, uh, joint, uh, uh, o- on, um, benefits of adding AI to, uh, to negotiations. And, uh, would like to, um, uh, would like to, uh, would like you to summarize very briefly for our audience the results of your research, um, uh, um, that you conducted among over 100, 120 executives with and or with- with- without, uh, uh, AI support. Uh, what happened with? What happened without? 

[Yadvinder]
Uh, what happens with is that, uh, if only one of the parties uses, uh, large language models at the time, because I was not able yet to develop agents, so I used large language models, and at the time, it was only ChatGPT. Uh, 

[Yadvinder]
LLM assistance provides really, really a lot of, uh, uh, competitive advantage. Uh, if only one of the two parties uses a... Used LLM, they achieved better, uh, better outcomes. Buyers gained 48% better than sellers. And sellers, if they used LLM, uh, gained 40% better than their buyer's counterpart. But, but I think the result is also very important for, uh, uh, integrative outcomes, because if you think about both parties using LLMs, uh, they achieve 84% better outcomes than, uh, the parties not using, compared to the parties not using LLMs. And this is, uh, uh, achieved even in low-trust settings. Why? Because, uh, uh, they can explore in parallel integrative solutions with LLM, then come to the table and discuss them, without having, uh, to, uh, exchange too much information. So it is very, very interesting because, uh, if both parties use LLM, uh, there are... There is more joint value, but also more fairness during the, the negotiation. So it, it really helps to have more integrative negotiations. LLM can be an enabler for integrative negotiations, according to the 

[Remi]
That's... That's super interesting, Edwinder. Maybe, uh, maybe let's dive into, um, into the mechanics or the logic be- behind, uh, behind this process. Yes? So, uh, why is that the case that, uh, using LLMs, uh, gets us more individually, but at the same time also, if it's symmetric, if it's applied symmetrically, uh, it also leads to creating more value? 

[Remi]
What is the, the, the mechanism that, uh, that is at work? 

[Yadvinder]
Uh, th- this is something that has to be explored much deeper, but my theory is the following. That according to, uh, traditional negotiation theory, in order to have integrative negotiation is you have to build trust, exchange information, and then you understand the interests of the other party, you come up with creative options. Uh, in this case, when both parties use LLM, they don't need to exchange information. They don't need to build trust in advance. But because they use LLM to prepare for the negotiation, they already explore the different scenarios, and they already have creative options that can be of value for other party. And when they come to the table, that creates a symmetry of knowledge that allows to have better outcomes, integrative outcomes, even if there is no trust between the parties, because you don't need to exchange information. In fact, there are higher joint gains even though there is reduced exchange of information, uh, when there is symmetric use of large language models. So the theory is, maybe you don't need to exchange information anymore, because you already did that with your large language model during the preparation phase. 

[Remi]
Are joint gains the ultimate goal of a negotiator? 

[Yadvinder]
[laughs] 

[Remi]
So in other words, uh, in other words, yes, it's nice. It's nice that, uh, using LLMs helps us- uh, help us, uh, get to the center of the Pareto efficiency frontier, or the, the midpoint of the efficiency frontier. But the question is, is that where we want to get?... what do you think? 

[Yadvinder]
Uh, I don't know if it is where we want to get but for sure, uh, what studies are- are telling us more and more is that if you don't use AI support then you will gain less than the other party, uh, that's for sure. Uh, but I like the fact that, uh, uh, you can have integrative negotiations even in, uh, if you don't establish trust earlier, eh, because most of the negotiations happen in a low-trust setting. So I think this is very, very important because allows us to have, uh, more integrative negotiations even in low-trust settings. I think this is very, very important. But also what I found in this, uh, uh, in this, uh, research that is still going on is that, uh, there is also increased fairness, uh, during the negotiation process. Uh, if both the parties know the other party is using AI, there is less deception, there is more transparency, and, uh, this leads to better outcomes at the end for both of them. Is this something we want? Uh, I think I- I like the fact that there is more transparency and more fairness in the negotiation. This is for sure. 

[Remi]
But, uh, coming back, uh, uh, coming back to, uh, to your question, do I think, uh, um, AI can build trust? I think your study shows that it can. Yes? Uh, and I think, uh, uh, even if it doesn't understand what trust is in terms of, uh, in terms of the feeling, in terms of the confidence, the belief, uh, right? Uh, it still can, uh, decode the components of trust building, and I think it can and it does, uh, use those components, uh, um, such as, for example, transparency in information sharing or higher tr- levels of transparency in information exchange such as kindness, uh, such as other elements that we typically associate with, uh, uh, trustworthy individuals, uh, or agents, yes, in that ca- in that particular case. Uh, so yes, I think, uh, uh, trust building is nothing inherently reserved for humans only. I think it can be, it can be, uh, it can be re- [sighs] it can be engineered as well, yes? When we, once we've understood a creation or development of a certain phenomenon, once we've understood the factors that influence it, we can build machines. Usually, we can build machines that, uh, uh, that can, that can, uh, that can do the same or similar.