Your Negotiation Agent Optimizes the Wrong Number

A distribution company I advise switched on an autonomous negotiation agent this spring for tail-spend renewals. The pitch was clean: point it at a few hundred small supplier contracts nobody has time to renegotiate, let it email back and forth, book the savings. It worked. On one packaging category it pulled unit price down 4%. The procurement lead forwarded the summary around. Everyone saw a win.

Three weeks later the cash forecast tightened and nobody could say why. I traced it back to the same renewals. To land that 4%, the agent had agreed to Net-20 terms in place of the Net-45 the company used to run. It did that across the book. Average days payable on the touched spend dropped from 41 to 28.

The scorecard and the cash line disagreed

Here's what that trade actually cost. The renewals covered about $6M of annual spend. Pulling days payable down 13 days moves roughly $214,000 of cash forward, permanently, off the balance sheet and into suppliers' hands earlier every cycle. The 4% price cut on the category it bragged about saved maybe $80,000 a year.

So the agent reported a win and quietly made the company's cash position worse. Not because it malfunctioned. Because it did exactly what it was scored on. These platforms optimize for headline price and reported savings. Pactum, one of the larger autonomous negotiation vendors, cites 3-7% additional value on tail spend as the number that sells the product. Payment terms aren't in that number. Neither is your cost of capital. The agent will give away a financing concession to win a pricing concession every time, because one shows up on its scorecard and the other doesn't.

Your negotiation agent doesn't know your revolver is at 70%. It knows it saved 4%. Those aren't the same decision, and only one of them made it into the forecast.

Give the agent a working-capital floor

The fix isn't to turn it off. Tail-spend renewals are real work no human was doing well. The fix is to stop letting price be the only thing the agent can see.

Set a terms floor before it negotiates anything. No supplier term shorter than Net-30 without a human sign-off, full stop. Then hand the agent a real trade rate: a day of payment terms is worth something specific in your business, so a price cut that shortens terms has to clear that hurdle, not just beat last year's price. If you carry a revolver, the number is your borrowing cost. If you sit on cash, it's your opportunity cost. Either way it exists, and the agent should be negotiating against it.

Most companies haven't written that number down anywhere, which is how the agent ended up optimizing without it. That gap was survivable when a person ran each renewal and felt the tension between price and cash. An agent feels nothing. It hits the metric you gave it.

Check what your agent traded away last quarter before you scale it to the next category. The savings are on the summary. The cash is somewhere in the forecast, waiting to be found.