For the COO / VP Supply Chain
You answer for OTIF, fill rate, attainment, and turns — and every one of them is set, day by day, in the allocation call, the expedite thread, and the spreadsheet a planner keeps between two systems. The A2go Decision Intelligence Platform (ADIP) puts agents on those daily decisions: they assemble the trade-offs with the reasoning attached, your team approves or edits, and the change is written back to the system that runs the operation.
The daily reality
Your ERP knows the orders. Planning has a forecast. The WMS knows what is actually on the shelf. None of them decides who gets shorted when a supplier confirms 60% of a purchase order this morning — so that decision gets made the way it always has: a 6 a.m. export, a 45-minute call, and an allocation spreadsheet that is stale before the meeting ends.
That is not a staffing problem, and more dashboards do not fix it. In the published Databricks customer story about A2go, one customer — 65 production facilities, 17 sales channels, more than 150,000 customers — started in exactly this state: analysts pulling data into Excel and circulating static recommendations that were stale on arrival.
The allocation call
The expedite thread
The retiring playbook
Tribal knowledge, captured
The knowledge that actually runs your operation is not in any system: which forecast to trust, which dock can't take a truck after 3 p.m., which contract customer to protect first when allocation gets tight. It decides more than any parameter in your ERP, and today it walks out the door with the people who hold it.
ADIP's Judgment Layer is where that knowledge stops being tribal. When a planner edits or rejects an agent's recommendation, the correction and its rationale are kept — not as a complaint log, but as policy the next recommendation respects. Constraints no system records get written once and enforced on every decision. New planners inherit twenty years of judgment on day one instead of apprenticing for it, and nothing leaves when someone retires.
The working day, changed
Not another screen to check. Agents watch the signals and assemble a decision package: what changed, the options they priced, the recommendation, and the reasoning. Your planners review it the way they would review a colleague's work — approve it, edit it, or reject it. Either way, write-back is handled: the purchase order re-dated in the ERP, the allocation adjusted in order management, the wave re-sequenced in the WMS, every change logged with its rationale. Autonomy is a dial you set per decision type — routine changes run within pre-agreed bounds; anything outside them waits for a person.
Worked examples
More decisions in the decision library · outcomes and provenance on Results · bringing your CFO along? The CFO page · security and governance questions: Trust
You already know which one it is — the meeting you dread, the spreadsheet that never dies, the number you explain every month. The survey takes about ten minutes and maps that decision: who touches it, what it costs to make it slowly, and what an agent-run version looks like.
Map your highest-pain decision