When the Other Side Isn’t Human—and You Don’t Know It
Imagine negotiating your salary.
You’ve done your research. You know your value. You’re ready to advocate for yourself.
But what you don’t know is this:
The offer you receive may have been shaped before the conversation ever started—by an AI system trained on thousands of compensation decisions, modeling how likely you are to accept, how far you might push back, and the minimum required to close the deal.
Companies like HireVue—previously scrutinized for using AI to assess candidate behavior and video interviews—and platforms such as Eightfold AI are already influencing hiring outcomes and compensation ranges before a hiring manager becomes involved.
You think you’re negotiating.
In reality, the negotiation may already be over.
This isn’t theoretical.
Companies are already using AI systems to negotiate commercial terms, set pricing, and influence financial outcomes—often without the other party realizing it.
The question is no longer whether AI is part of negotiation.
Now the real question is:
What does it mean to negotiate fairly when one side is an algorithm?
You May Already Be Negotiating With AI
AI systems are already participating in negotiations across multiple domains:
- Procurement: Companies like Walmart use platforms such as Pactum AI to autonomously negotiate supplier contracts—settling pricing, payment terms, and discounts directly with human vendors.
- Healthcare: Insurers increasingly rely on algorithmic systems to determine claim payments and influence reimbursement decisions. Investigative reporting and litigation in 2023–2024 alleged that AI-driven systems were used to deny claims at scale, in some cases with limited human review.
- Hiring and compensation: AI tools shape salary bands, candidate scoring, and initial offers—often setting the parameters of negotiation before it begins.
- Digital markets: Pricing algorithms dynamically adjust offers based on demand, behavior, and competitive signals.
In many of these cases, people don’t realize they are interacting with an optimized system.
They assume they are negotiating with a person.
But the “other side” may already know:
- how likely they are to accept
- how much resistance they will offer
- and where their breaking point is
This is not just a technological shift.
It is a redefinition of the negotiation itself.
The Ethical Tension
Informed Participation
Negotiation assumes both parties know what kind of interaction they’re having.
That assumption is starting to break.
If one side is using an AI system designed to optimize outcomes—and the other side is not aware—then the negotiation is no longer fully informed.
At a minimum, this raises a basic ethical question:
Is it fair to negotiate with someone who doesn’t know they are negotiating against a machine?
Regulators are beginning to take a position.
Under the EU AI Act (2024), AI systems used in employment and insurance contexts are classified as high-risk, with specific transparency obligations—particularly when decisions materially affect individuals.
The direction is clear:
Disclosure is not just good practice.
It is becoming an expectation.
Power Asymmetry
Negotiation has always involved unequal power.
AI changes the magnitude.
An AI system can:
- analyze thousands of comparable deals
- simulate how the other side might respond
- optimize strategies in real time
An individual cannot.
We’ve seen this dynamic before.
For example, in financial markets, high-frequency trading created such big speed differences that regulators had to introduce safeguards like circuit breakers to keep markets stable.
AI-driven negotiation may be creating a similar asymmetry—but without equivalent guardrails.
At some point, negotiation stops being a discussion.
It becomes a system optimizing against a human.
The Illusion of Fairness
AI systems often give the appearance of neutrality.
Offers feel data-driven. Decisions feel consistent. Outcomes feel justified.
But that sense of objectivity can be misleading.
The system is not neutral.
It is optimized.
Optimized for:
- conversion
- cost reduction
- margin improvement
- settlement speed
Not fairness.
And because the decision is framed as “the data,” it becomes harder to question.
Consider a claimant receiving an insurance settlement determined by an algorithm, presented as an objective calculation—with no clear explanation and no human to challenge.
Or consider a job candidate who gets a final offer that seems fixed and backed by “market data,” when in reality it is designed to secure acceptance at the lowest possible cost.
A human negotiator can be challenged.
An algorithm feels authoritative.
That makes the imbalance less visible—and more difficult to contest.
Collusion Without Intent
One of the most striking ethical challenges is not intentional misconduct—but emergent behavior.
A study by Calvano et al. (2020) demonstrated that independent pricing algorithms, using reinforcement learning, were able to learn collusive-like behavior in a simulated duopoly.
They were not programmed to collude.
They did not communicate.
Yet over time, they discovered that maintaining higher prices was more profitable—and converged on that outcome.
The coordination emerged purely through learning.
No human instructed the system to behave this way.
No explicit agreement was made.
And yet the outcome resembled classic price collusion—higher prices, reduced competition, and worse outcomes for consumers.
This reveals something deeper:
AI systems can produce harmful outcomes without intent, instruction, or visibility.
And when that happens, responsibility becomes difficult to assign.
No one designed the behavior.
But the impact is real.
Human Dignity
At its core, negotiation is not just a transaction.
It is a human interaction.
It involves judgment, context, empathy—and sometimes, compromise.
When those elements are replaced by optimized systems, something changes.
People are no longer negotiating as individuals.
They are being modeled, predicted, and influenced.
That shift matters.
Because when outcomes are determined by how well a system can predict your behavior, the negotiation is no longer about mutual agreement.
It is about strategic extraction.
A salary offer calibrated to the minimum you will accept.
An insurance settlement timed for when claimants are most likely to give in.
A supplier price set just below a vendor’s walk-away point.
That is what optimization looks like in practice.
What Should We Do About It?
Before organizations decide how to use AI in negotiation, they need to decide what they believe is acceptable.
For executives, that starts with asking the right questions:
Executive Checklist
- Would we be comfortable disclosing to the other party that they are negotiating with an AI system?
- Are we optimizing for efficiency alone—or are we considering fairness and long-term trust?
- Can we explain how a negotiation outcome was reached in human terms?
- Have we defined limits on how far the system can push outcomes against individuals?
- Are we monitoring for unintended behaviors, including exploitative or collusive outcomes?
- Would this approach withstand scrutiny from regulators, courts, or the public?
These are not compliance questions.
They are leadership decisions.
Executive Takeaway
AI is changing the nature of negotiation.
What was once a human exchange is becoming a system-driven interaction—faster, more precise, and increasingly invisible.
But when one side is an algorithm and the other is not, the question is no longer just who got the better deal.
It is whether the deal was fair in the first place.
Because if one party understands the system—and the other does not—
it’s no longer a negotiation.
It’s an optimization problem.
And one side is being solved.
When AI negotiates, it is still making human decisions—just without the human in the room.
Sources
Calvano, E., Calzolari, G., Denicolò, V., & Pastorello, S. (2020). Artificial Intelligence, Algorithmic Pricing, and Collusion. American Economic Review.
European Union. (2024). EU Artificial Intelligence Act.
OECD. (2023). Artificial Intelligence in Markets and Competition Policy.
U.S. Federal Trade Commission (FTC). (2023). Protecting the Public from Deceptive AI Claims.
https://www.ftc.gov/business-guidance/blog/2023/02/keep-your-ai-claims-check
Pactum AI (2023–2024). Autonomous Negotiation Platforms in Procurement (e.g., Walmart use cases).
Investigative reporting (2023–2024). AI use in healthcare claims adjudication and related litigation (e.g., UnitedHealth Group cases).
