Two things it does better
Most automated negotiators either concede on a fixed timetable or simply push for their own best outcome, and often end with value left on the table or no deal at all. AI assistants negotiate fluently but unreliably: some hold out too hard, others give too much away.
Makes AI assistants negotiate better
The engine sits alongside an AI assistant and tells it what to offer and when to accept. The assistant keeps the conversation; the engine keeps the deal on track.
Closes more deals without giving value away
It works out during the negotiation what matters most to the other side, and uses that to find trades where both parties gain, so fewer deals fail and its own side keeps its ground.
Also built in
Explains every decision
Each move comes with a plain-English reason, so buyers, managers and auditors can see why it acted as it did.
Remembers who it deals with
Learns how each supplier negotiates and carries that into the next meeting, while staying alert to change.
A style you can set
From cooperative to competitive, to suit the market and the relationship.
Runs on its own or with AI
Negotiates by itself, advises a person, or guides an AI assistant, as a hosted service or built into your product.
What it could mean for you
Enter your own volumes. The figures use the engine's measured test results; your results will depend on your negotiations.
Basis: as the buyer in 200 simulated supplier negotiations, the engine closed 7–10 percentage points more deals than a tough negotiator (92% against 76–86%), while also gaining value for the buyer.
Basis: an AI assistant advised by the engine closed 98% of negotiations against 69% on its own (29 more in every 100) and reached 78% of the best joint deal against 41%; in earlier tests advised assistants generated up to 26% fewer words.
Results so far
From thousands of test negotiations. “Combined value” is how good the deal is for both parties together, as a share of the best deal possible.
| Test | Standard approach or AI alone | Agreement Engine |
|---|---|---|
| Advising an AI assistant, on the same negotiations | Assistant alone: 41% combined value, 69% of deals closed | 78% combined value, 98% of deals closed |
| As the buyer in 200 simulated supplier negotiations (price, payment terms, volume, delivery, contract length) | Tough negotiator: 76–86% deals. The 2026 competition winners, as buyers: 30–64% | 92% deals, more value for the buyer, and suppliers also better off |
| Against standard negotiating programs, on 114 real negotiation scenarios | 81% combined value | 88–89% combined value, and a deal in every negotiation |
| Against 11 prize-winning programs from the international negotiating-agents competition (2010–2019) | 94% deals, 75% combined value | 99% deals, 79% combined value, the same value for its own side |
| Against today's leading AI models, on the same scenarios | 80–85% combined value | Within 1–3 points of them with no AI model running |
Results are from simulated and recorded negotiation scenarios. A study with real people negotiating against it is planned.
Where it can be used
- Purchasing and supplier deals: price, payment terms, delivery and volume with many suppliers at once, with fewer failed deals
- AI assistants acting for people: booking, buying or scheduling on someone's behalf, with better results for everyone involved
- Business-to-business AI agents: software agents negotiating with other software agents
- Marketplaces and sales: helping buyers and sellers settle on terms both accept
- Dispute resolution: proposing settlements both parties can live with
Get in touch
For a demonstration, evaluation access, licensing or a partnership, send a message and I will reply by email, usually within two working days.
Julian David Hunter
julian@theagreementengine.com
LinkedIn profile
Evaluate it yourself
Patent pending: four UK patent applications filed in October 2026. Available for licensing or acquisition, as a hosted service or for building into products. Prospective partners can test it on a private evaluation service, under a confidentiality agreement.