Priced alternatives, not a red dot

What the red dot leaves out

The red dot is the signature move of supply chain software: a KPI turns color, a threshold trips, an exception lands in a queue. It names a problem — a late supplier, a short position, a promise at risk — and that is where it stops. The dot carries no options, no costs, no constraints. It prices nothing.

So the expensive work still happens — it just happens off-screen. A planner opens window after window, reconstructs the situation, imagines two or three ways out, guesses at what each would cost, and picks one under time pressure. The software gets credit for the insight. The person absorbs the decision.

What a priced alternative set is

When an ADIP agent opens a decision, it returns the whole choice, not just the headline. A priced alternative set has four parts:

• The options. The credible ways through — expedite, reallocate, re-promise, split the difference — not just the obvious one.
• The price on each. What every option costs in freight, penalties, margin, and strain — including the option of doing nothing, which is never free.
• The constraints, already applied. Contractual commitments, capacity limits, and the judgment your planners have encoded. Options that violate them don't appear dressed up as bargains.
• A ranked recommendation with its reasoning. Which option the agent proposes, and why — written so you can argue with it.

Same fill, different exposure

Consider the classic allocation squeeze: more committed orders than available supply. Every strategy on the table ships the same total quantity — there is no more product to ship. What changes between them is who gets shorted, and what each version of that choice costs in penalties, contract exposure, and relationships.

A red dot renders that situation as “inventory below commitment.” A priced alternative set renders it as it actually is: several defensible ways to allocate, each with a different bill attached, one of them cheapest once the whole bill is counted. Same facts. Only one of the two is a decision.

That shape repeats everywhere: the carrier tender nobody wants, the production slot two orders are fighting over, the forecast that just left its band. Different families of signal, same anatomy underneath — a constrained choice among costed options, waiting for someone to see it whole.

The scarce thing isn't detection

Plenty of software can detect. That part is everywhere — which is exactly why alerts get tuned out. What stays scarce is the assembled decision: options, prices, constraints, and a defensible recommendation, delivered while the cheap options still exist.

A recommendation without reasoning is a black box asking for trust. A recommendation with its alternatives priced beside it is a colleague showing their work — something you can interrogate, adjust, or overrule.

That division of labor is the whole design. Agents assemble and price; your people judge — with the costs in front of them, not reconstructed after the fact. The next step of the loop is where that judgment gets applied.