Why the Future of AI Negotiation Isn't a Chatbot

The future of AI-driven procurement negotiation isn't a chat interface where a digital assistant talks to a supplier the way a human would. It's protocol-based: agents exchanging rich, unambiguous data directly with other agents, pooling demand and processing combinatorial bids at a speed and precision human conversation can't match.

Key takeaways

• Machine-to-machine negotiation outperforms chat-based interfaces because it removes the ambiguity and slowness of human language
• AI agents can form "implicit consortia," pooling buyer demand for cost synergies without buyers ever knowing who else is involved
• Combinatorial bidding lets agents surface cost synergies across bundled needs that a human negotiator would never see
• The goal isn't to make AI act like a human negotiator; it's to let AI optimize in ways conversation never could

There is a common misconception that the "Agentic AI" revolution in procurement will look like a more sophisticated version of ChatGPT—a digital assistant you "talk" to so it can "talk" to a supplier. For an interim period there will certainly be elements of this as we make progress but this will be short-lived.

I believe that is fundamentally wrong because it fails to recognise something very obvious; sales organizations are using AI too, and furthermore, so too are other buying organizations. This matters when it comes to systems evolution and machine to machine communications. Chat based interactions are slow, inefficient and ambiguous. They are important for human to machine communication but unsuited to machine to machine communications. B2B trade values speed and efficiency in negotiations and as machines take charge of tactical and tail spend within Procurement, efficient markets will evolve quickly in a world of intelligent systems. These negotiations will benefit from very rich data and powerful optimization to maximize efficiency and speed.

If we limit AI to conversational interfaces between machines, we are bottlenecking its true potential with the limitations of human language. The real breakthrough for B2B trade won't be chat-based; it will be protocol-based. We are moving toward a world where Sourcing Agents (and consortia of Sourcing Agents) and AI Sales Agents negotiate via workflows that are data rich and involve models for optimizing economic efficiency. When we remove the constraints around concise human language and its inherent ambiguities, we see lots of potential to achieve much greater efficiency and speed.

The Rise of "Implicit Consortia" & Combinatorial Power

What is an "implicit consortia" in AI-driven sourcing?

Historically, buying consortia (like Sourcing Alliances) were manual, infrequent, and administratively heavy. They were reserved for massive, pooled spend categories and only operated by the top sourcing teams at public businesses.

In an agentic world, pooling happens at the speed of light. AI agents can form "implicit consortia" where buyers benefit from aggregated demand without ever knowing the other parties involved. The magic happens when agents share unambiguous, rich data rather than conversational pleasantries. Consider this classic logistics scenario now coming to life:

• Buyer A (Retailer): Needs a truck from Chicago to Detroit.
• Buyer B (Manufacturer): Needs a truck from Detroit to Chicago.
• The AI Bidding Agent: Recognizes the cost synergy of a continuous move. It offers a "package discount" to both, winning both items.

Both organizations win on price. The carrier wins on asset utilization. And the two buyers never even spoke to one another—the agents identified the synergy through a combinatorial exchange that handled the complexity behind the scenes. These synergies exist in all kinds of other spend categories, from chemicals to Facilities Management and direct materials.

From Conversation to Combinatorial Logic

How does combinatorial bidding work in procurement negotiations?

Fifteen years ago, AI research literature conceived of these combinatorial exchanges. Today, the compute power and agentic frameworks are finally here to make them a reality.

If your AI strategy is focused on building a bot that can "argue" for a 5% discount in a chat window, you are missing the forest for the trees. The real value lies in multi-agent systems that can:

• Pool demand dynamically across global networks.
• Process combinatorial bids (discounts contingent on winning specific "bundles").
• Optimize for cost synergies that humans can't even see.
• Learn from outcomes: biasing in favor of high performing suppliers.

This is why Keelvar built a Sourcing Optimizer as a workflow engine; not just to support the large strategic sourcing (that will remain human operated) but to empower buyside and sellside agents to express rich synergies and efficiencies in complex supply chains.

The Bottom Line

What should procurement leaders actually be optimizing their AI strategy for?

We need to stop trying to make AI act like a human negotiator, it may well act as a means to an end, but ultimately we can expect AI to remove those shackles and embraces optimization for a step change in the results it produces and not merely for operational efficiency within Procurement. The most effective negotiations of the next decade won't be a back-and-forth dialogue. They will be a high-speed exchange of rich data using multi-round protocols that lean on optimization and explore scenarios dynamically —a richer, more transparent way of doing business that yields benefits for both sides.

Is your agentic strategy for tactical and tail spend built for conversation, or is it built for the combinatorial exchange?

Chat to Keelvar to learn more