Programming AI Agent Values: The Strategic Choice Behind Every AI Negotiation

When deploying AI negotiation assistants, organizations face a fundamental question: should your agent extract maximum value or seek fair, mutually beneficial outcomes? This isn't a technical decision—it's philosophical, reflecting your company's approach to relationships and value creation.

In this episode with Prof. Remi Smolinski, we explore how sales teams use integrative approaches to move beyond price discussions, while procurement requires sophisticated routing using the Kraljic matrix to match suppliers with appropriate strategies. We also examine how agents can adapt strategies in real-time based on counterpart behavior.

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]
In the implementation phase of the, um, AI bots, the principal has to answer a fundamental question is, how do I want this bot to interact or this agent to interact with my customers or with potential, uh, potential sellers? And there's a wide range or wide spectrum of potential answers to it, from, I want it to exert every single penny out of every salesperson, yes, to find a Pareto efficient outcome and share the benefits, uh, uh, equally or similar or fairly. Yes. Yadvinder, what's your experience while dealing with, uh, com- companies that you consult? Because when you ask them this question, 

[Remi]
they should pause and give you an answer. What is their answer usually? Is it do they push for fairness or do they usually try to use this tool as an, uh, ano- as, as, as a different way of exerting their competitive advantage? 

[Yadvinder]
It, it depends on the role, uh, they're asking me to create th- the systems. If it's a sales rep, they want, uh, uh, a more integrative negotiation. So something like, "Let's create value together," uh, because that moves them away from a transitional... Uh, transactional, sorry, uh, negotiation where they, they just have to give discounts. But if you talk to buyers, then, uh, there is a range of possibilities there. Uh, if you're talking about procurement or purchasing departments, it depends on, uh, the, the purchasing, uh, director or procuring director. Some of them are really only on price. And, uh, and it's not even fun then to create these workflows, uh, because, uh, they only have to achieve one, uh, one objective. Uh, it's much more fun when they're looking for partnership. So what we do is usually based on the Kraljic matrix. We decide, uh, how to deal with different types of, uh, suppliers, depending on their position on Kraljic. And so some of them will be really price-oriented negotiations. Some of them will be more towards partnerships. And this is much more fun, because, uh, you have to create a first, uh, buyer agent that is able to discern which kind of supplier they're meeting and based on all the knowledge they have, to adopt the strategy and tell the strategy to the negotiator agent. So it is very, very interesting. Uh, fine-tuning is not easy, but still it is, uh, it is something you can do based on, uh, the negotiation, uh, uh, partner you are meeting. Uh, but this is very interesting because usually buyers tend to go towards price, uh, in all circumstances. Uh, so I think it is very interesting because it allows them to have more, uh, more strategies, different tactics in a negotiation. 

[Remi]
That's absolutely amazing, uh, what, what we're hearing here from Yadvinder, from you, uh, here, in terms of, uh, in terms of the, uh, war room or boardroom stories, uh, uh, uh, in terms of implementations, uh, implementations and objective, uh, objectives that are set for, for, for the AI agents. And I was, um, and I was thinking about, uh, about the case of Walmart, as Walmart negotiating probably using Pactum AI's, uh, uh, pack- software package, uh, with hundreds or thousands egg suppliers. As, uh, um... Do you think... Uh, what is your... What would you speculate? Do you think their objective function would be to create potential partnerships or would it be more on, um, on, uh, on, um, claiming, claiming value in terms of, uh, the price of those eggs? 

[Yadvinder]
Uh, uh, there was also some kind of shortage of eggs in, uh- 

[Remi]
[laughs] Yeah 

[Yadvinder]
... in the US lately. So I think, I think if you think about- 

[Remi]
That's probably, it was probably caused by Pactum's AI implementation at Walmart. [laughs] 

[Yadvinder]
Maybe, yes. But if you think about, uh, the shortage of eggs, then, uh, certainty of supply is more important than price. And then- 

[Remi]
Sure 

[Yadvinder]
... maybe I would gear Pactum in that negotiation towards that aim more than the final price. But for sure, if you're Walmart and you're using Pactum, I would assume that in most of the negotiations, price becomes the first priority, for sure. 

[Remi]
So key question is yet, Yadvinder, what do you think? Are human, are humans better than machines when it comes to negotiating today and potentially in the future? 

[Yadvinder]
[laughs] Uh, that, that, that's a very mean question. 

[Remi]
I know, I know. [laughs] 

[Yadvinder]
And, uh, I'll, I have a, a very straightforward, uh, response. I think today, uh, at the level of technology we have, um, with the agents we are able to develop, uh, or assistants or autonomous agents, I think, 

[Yadvinder]
uh, that autonomous agents and augmenting assistants are better in preparing for a negotiation and negotiating than the average human. 

[Yadvinder]
I don't think they are better than the best negotiators we find. You know, w- we teach, uh, in MBAs, executive MBAs and everything, and, uh, finding good negotiators is not easy. Uh, so I really think that because of... if we train them well, if we have, uh, proper system instructions, uh, and they become reliable, they can be better negotiators than the average of human negotiators, as of today. Even today. Even today. Uh, I'm, I'm really sure of that. And, uh, and I think they can also be as good as the best of the negotiators. 

[Remi]
That's amazing because, uh, uh, I might have mentioned, uh, that we regularly run negotiation competitions. Uh, and recently someone asked me whether, uh, whether we would allow, uh, an, an AI agent to enter the competition. And I immediately, uh, answered, "Of course."Yes, uh, so there's an open call out there, yes, for... to all of those, uh, who, um, design negotiation agents, or believe, uh, that they can design negotiation agents who are... who can beat some of the world's best negotiators. Uh, guys, join our competition. All right, uh, so, um, before we answer... Uh, before we answer Shane's question, um, I was going to, uh, I was going to ask you, uh, Yadvinder about, uh, about, uh, the platform that you use, uh, yourself, uh, uh, in your work, uh, to develop... uh, to develop agents. Tell us a little bit more how it looks like, uh, how it looks like, uh, from the kitchen. 

[Yadvinder]
Uh, I- I'm using TypingMind. It is a platform where you can use all large language models and you can set the temperature, all the parameters, uh, to develop the agents, Uh, it's a great tool. You're only charged once for your lifelong experience with the... with the platform, and then, uh, uh, you're charged by use. So some- You... You understand how tokens work and everything, but you're able to really develop very strong and very, uh, very performant agents there. And then I move to Cassidy. Cassidy is a platform where you can have different agent interacting among them, and this is where I build the workflows. I don't build them in Cassidy because it would be more... too mu- too expensive, uh, this is the reason. So I have a platform where I build them, that is TypingMind, and then I move to Cassidy where I, uh, develop them and develop the workflows. I sometimes will also use NA10 that is, uh, an automation platform, but this is more for automations than, uh, for augmenting. 

[Remi]
My last question. My last question, uh, uh, for you, uh, today is 

[Remi]
a request for advice. A request for advice, Yadvinder, um, uh, we... It seems that, uh, um, we're at a crossroads. Yes? A technology has entered the stage and, uh, it seems to be changing, or ha- at least, uh, it seems to have a huge impact on the way we negotiate, uh, on... Uh, it also replaces us in certain types of... or has the capacity of replacing us in certain types of negotiation. Um, so what shall we do, Yadvinder? What shall we do? 

[Yadvinder]
Uh, something that I s- I'm saying in all my classes is, uh, you have to combine and integrate negotiation expertise with, uh, AI literacy, for sure. Uh, because studies demonstrate that using AI you improve your outcomes. Uh, you become a better negotiator. So I really think you have to integrate them, and being only a negotiation expert is not enough today, as it is only being an AI expert. You have to combine AI expertise with domain expertise.