Since yesterday: Alpro EU moved its delivery promise a third time, the Retail Store's oat milk fell to 5 days of cover, and 4 suppliers answered your price requests. Everything else is stable.
12 weeks of real sales, then 8 weeks of forecast with its honesty band. The dip in cover is why decision no. 1 exists.
Each card says what happened, why, what boynt recommends and the alternatives it already priced. Nothing is sent without a signature.
5 days of cover left; next week needs 540 cartons (range 480–620, 94% confidence). Demand ran above forecast two weeks straight while the warehouse holds unused surplus.
14 customer orders exposed. Alpro EU's real delivery time is 12 days — double its promise — and its on-time score fell to 61%. You buy 28% of your Plant Milks from this one supplier; best practice is under 20%.
21 days of shelf life left (estimated until lot tracking connects) and expected sales won't clear the quantity. A seasonal order landed just as café season slowed.
Every agent works under your team's permissions: agents propose, humans decide. boynt reads a copy of your ERP hourly and writes nothing back.
Every product, every location, nothing summarised away. Type to find anything; rows with an action already have a priced recommendation.
| Product | Location | State | Cover | Next week (P50) | boynt's action |
|---|
"On time" means promise versus actual receipt from your own records — not what the supplier claims.
boynt's read: qualify alternates for the 17 single-source items and open a competition before the next cycle.
In 3 of the last 4 Alpro delays, the alternate closed the risk faster even when slightly dearer.
| Supplier | Category | On-time (12m) | Real lead (P90) | Spend / yr | Price 6-mo trend | boynt score |
|---|
Approved decisions become drafted POs and messages. Failed and late sends are shown too — the log is append-only.
Supplier silent after the 3rd promise change — escalates automatically at 16:00; the alternate stays open.
Sent from an approved template; the reply returns to the same case automatically.
Approved by the Operations Director — truck scheduled tomorrow 09:00.
GRN matched 100%, on time — case closed, outcome moved to Memory.
4 replies in — incoming prices are proposals; nothing applies before your approval.
July, computed from your ERP after each decision ran its course. The section your accountant will like.
The internal transfer closed stockout risk without an urgent purchase — third success in a row. Rule updated: trigger at a cover gap over 3 days.
Paying to expedite didn't change the arrival date. That measured outcome is why today's expedite option carries lower confidence.
A structured sell-down saved 7 margin points versus an immediate deep discount — now the default play for expiry cases.
Permissions are sentences, not checkboxes — what you read here is exactly what the system enforces. Threshold changes apply on the next sync; nothing needs a deploy.
Every agent acts with the permission of the person it represents — never beyond. Financial figures outside a role's visibility are omitted server-side, never just hidden.
Sees everything, all locations, full financials. No approval ceiling. The only person who can ratify exceptions — and every exception is recorded by name.
Approves purchase orders up to 100K AED and liquidations up to 30K. Cannot approve transfers — those route to the Operations Director automatically. His agent qualified for Level-2 autonomy at 84% acceptance.
Transfers between locations are his exclusive decision up to 50K AED — above that goes to the GM. Only this role sees the whole network's surplus-and-shortage picture.
Full financials, margins and collections — with no operational approval buttons: she can see every number but cannot approve a purchase or a transfer.
His location only, quantities without amounts — every financial figure shows "value outside your permission" instead of a number.
His warehouse only. Transfer authority is a hard zero — the transfer button reads "request from the Operations Director".
Autonomy is earned by measured performance, never assumed. Anything above a limit escalates — it never self-executes.
Tenant-scoped — a change takes effect on the next sync, with no deploy.
boynt reads a copy of your ERP every hour and writes nothing back. The score rises as the two missing feeds connect.
Channels carry decisions — they never create authority. Every message comes from an approved template and returns to its case.
The Copilot answers from your own numbers only — it never invents a figure, and every answer carries its source.
Identity and locale for this tenant.
These are product law, not preferences.
Measured from your receipts — the primary first, then the best alternates in its category.
Source: Supplier + PurchaseFact, 12 months · lead times assumed until receipt dates connect