Agent-to-Agent negotiation is an AI buyer agent conducting a real, multi-round commercial negotiation with supplier-side systems, without a person driving each exchange. In METIS it runs against live market benchmarks and stops inside the approval thresholds you set.
How is it different from a chatbot?
A chatbot answers questions about a purchase. A negotiating agent changes the price of one. The distinction is authority and state: our buyer agent holds a mandate, a benchmark and a walk-away point, and it carries those across several rounds with several suppliers at once.
What happens in a round?
- The agent collects valid opening quotes and averages them into a Market Baseline
- It counters each supplier against that baseline, plus terms and lead time
- Suppliers revise, and the agent plays the strongest offers against each other
- A best-and-final round closes it, and the winner is scored and recommended
Every round is logged. The recommendation that lands on the approver's screen shows the baseline, the final price and the gap between them.
Does a human still approve the purchase?
Yes, always. The platform is built so a purchase order issues only after an authorised approver signs it off. The agent negotiates, scores and recommends; the commitment stays with your team.
What does it actually save?
On tail spend the saving comes from two places: negotiating buys that were never negotiated before, and doing it without consuming buyer hours. Both matter, because the reason tail spend goes unmanaged is not that nobody could negotiate it, it is that nobody had time to.