NEGOTIATION AGENTS

Rehearse a negotiation before it counts.

Negotiation Agents is a simulation for training negotiators. AI agents take the other seats at the table, each with its own cultural background, strategy and red lines. Sarah Christoyannis brought the idea; I refined it and built the engine.

Watch the 70-second demo → · Discuss a simulation →

Why simulate

Role-play exercises with professional trainers are expensive and rarely repeated. Colleagues tend to hold back, and no trainer can keep up a cultural persona while tracking a dozen constraints at once.

Negotiation reflexes come from practice. A simulation makes that practice available on demand, as often as needed and at low marginal cost.

How it works

Issue by issue

The engine sequences the agenda: build trust on easier points first, then move to the hard trade-offs.

Red lines that hold

Non-negotiable positions are enforced in code. An agent cannot be talked past a hard limit.

AI and people together

Human participants join via Signal, web, email or terminal, alongside the AI agents, in one session.

Traceable sessions

Claims are checked against external sources during the session; hallucinations (confabulations) are flagged. Every session leaves a timestamped transcript, structured logs and JSON state.

Negotiator profiles

41 behavioural archetypes, grounded in Hofstede, the Big Five and the Harvard Program on Negotiation, can be combined in layers: culture, personality and strategy. Examples range from the consensus-driven Diplomat and the competitive Shark to the Creative Solver and the methodical Guardian.

Who it is for

It was designed with the Brussels bubble in mind: diplomatic training, EU and public-affairs teams, and associations preparing for difficult talks. The same engine works for any negotiation with several parties and clear positions.

Which negotiation do you want to rehearse?

Negotiation Agents is in early access. The project site shows the scenario builder and a live negotiation in a short walkthrough.

Project site with demo → · Discuss a simulation →