Every tile here is either quiet — a verified, timestamped "all clear" — or a decision card: the situation, a prediction with its evidence and track record, two or three options with simulated consequences, a recommendation, a decide-by clock, and the cost of deciding nothing. The models run the loop — sense, predict, simulate, surface, verify. The human spends judgment only where judgment changes the outcome. That's how a mid-market company gets a Fortune-100 operating brain without the headcount.
Continuity: this is the week after the operator showcase. PO-88214 shipped complete and paid on time. The proofer drift ticket from chapter 5? It grew into a maintenance prediction — you'll decide what to do about it.
| Decision | Owner | Decide by | State |
|---|---|---|---|
| Cream-cheese stockout before Thursday's pumpkin run predicted gap 1,300 lb · 3 options simulated | Ops / Supply | Today 14:00 | PENDING |
| Proofer #2 maintenance window exceedance predicted Nov 19 ±2d · PM vs defer vs run | Ops / Maint | Thu 12:00 | PENDING |
| Hartwell asks +3,000 cases for 11/20 capacity simulated · margin holds at 31.8% with OT | Sales | Wed EOD | PENDING |
| What ran itself | Level | Evidence |
|---|---|---|
| Film stock reorder, $412 (consumables ≤ $500 policy) | L3 | PO sent 02:14 · confirmation received 06:50 |
| 2 EDI UOM exceptions resolved as drafts; CSR approved 07:40 | L2 | suggested fix matched catalogue rules |
| Missing invoice caught: DN-00311 shipped 11/7, no invoice — draft created, pending Finance review | L2 | 3-day unbilled threshold tripped |
| 14 price-list consistency checks passed | L4 | read-only verifications, logged |
Thursday's pumpkin run (WO-2026-1051 · 1,800 CS · feeds PO-88401 + holiday stock) is the week's pressure point — its material gate is in the Supply lens.
| Run | Material | Need | On hand + on order | Gate |
|---|---|---|---|---|
| WO-1051 · Thu | Cream cheese (Lakeland) | 2,400 lb | 1,100 + 0 | GAP 1,300 lb |
| WO-1051 · Thu | Pumpkin purée (Valley Fruit) | 900 lb | 1,420 + 0 | CLEAR (watch: COA drift — Quality) |
| WO-1053 · Fri | Graham base | 1,600 lb | 2,210 + 0 | CLEAR |
| all lines | Pack film | — | reorder in transit (auto, L3) | CLEAR |
| Asset | Signal | State |
|---|---|---|
| Proofer #2 | temp drift +0.4°F/wk | PREDICTED EXCEEDANCE NOV 19±2 |
| Line 1 wrapper | vibration baseline | NORMAL |
| Freezer bank A | compressor cycles | NORMAL · drilled Sun ✓ |
| Label printer L1-P2 | vendor EOL notice | REPLACEMENT SCOPED |
| Order | Customer | Where it is | Promise | Forward risk |
|---|---|---|---|---|
| PO-88377 | Midwest Grocers | In production (L2) → pack Thu | 11/12 | ON PACE |
| PO-88401 | Midwest Grocers | Scheduled · pumpkin Thu run | 11/16 | AT RISK — stockout decision pending |
| PO-88455 | Hartwell Grocery | Confirmed · materials clear | 11/18 | ON PACE |
| PO-88460 | Prairie Fresh | Intake accepted · 14/14 lines clean | 11/20 | ON PACE |
OTIF trailing: 98.4% · streak 14 weeks · every row drills to the operator-level screens — the EDI inbox, the work order, the pallets — on the floor showcase.
| Signal | Read | State |
|---|---|---|
| MGC ordering 18% under seasonal baseline, 3 weeks | no holiday promo booked this year vs last — churn signal or timing? account review pack prepared | WATCHING |
| 7 growth accounts, zero 6ct orders | the target list from the analytics portal — 3 converted since Oct | WORKING |
| Pumpkin velocity vs last season | +9% same-week · supports holiday build decision (Finance) | HEALTHY |
| Card | Simulation | State |
|---|---|---|
| Holiday inventory build build 4,000 CS pumpkin ahead vs produce-to-order | ties up $61K for ~5 wks · protects peak OTIF · storage clear | MODELING — needs demand commit from Sales |
| EDI VAN cancellation notice replacement portal live since July | notice by 11/15 saves $31K/yr · zero switch risk remaining | RECOMMENDED — awaiting signature |
| Formulation tool renewal (1/15) $84K/yr trajectory · data fully extracted to own DB | negotiate from strength or exit · both paths costed | QUEUED — Dec decision |
| Exception | How it surfaced | State |
|---|---|---|
| DN-00311 shipped 11/7 — no invoice raised | 3-day unbilled threshold · draft auto-created overnight (L2) | PENDING YOUR REVIEW |
| Freight invoice 4% over contracted rate | rate-card check on intake | DISPUTED · credit requested |
| Early-pay discount, flour vendor | 2/10 net-30 scanned = 36.7% APR-equivalent · taken Mon | DECIDED ✓ · saved $488 |
| Cluster | Trace | State |
|---|---|---|
| 3 texture complaints, same week | all map to B-1044-02 · bake-profile variance 11/02 · 0 cases left in DC · containment complete | CONTAINED · CAPA OPEN |
| Label date-format query (1) | retailer-specific template confirmed correct | CLOSED |
The cluster was surfaced by the system reading complaints against lot genealogy — nobody searched. The question arrived already answered.
Click WHY? on any card: data lineage (which flows, freshness timestamps), the model's name, and its graded track record (84% supply · 91% maintenance). Trust is earned on screen, never assumed. A prediction without evidence doesn't ship.
Options arrive with consequences already run against the live schedule, materials, labor and cash — so the leader compares futures, not guesses. The cost of deciding nothing is always printed. Urgency can't skip the math.
L1 suggests · L2 drafts for approval · L3 acts within pre-approved limits ($412 film reorder) · L4 only for read-only checks. Judgment calls stay human. Every autonomous act lands in the ledger for audit. Default is always the lowest level that works.
Every decision is appended with what was predicted; outcomes are reconciled weekly. Last week: 11 decisions, 9 graded correct, 2 pending. Predictors that decay get retrained or retired; decisions that keep beating the recommendation teach the recommender. The institution learns on purpose.
Everything here reads the governed datalake, the MES, the ERP and the monitoring fleet shown in the operator showcase — 18 flows, 13,847 live metrics, freshness-checked. The data layer is LIVE in production patterns today; this decision layer is DESIGNED — the layer we build per client on top of it.
Today you played three decisions: a stockout caught three days early, a machine fixed before it failed, and a promise made against live capacity. In the company we run, the data layer beneath every one of those moments already exists — the thread, the loop, the governed lake. The decision layer is what we build next, per client, seat by seat.
See the same company from the floor: showcase.layofthelan.com — one order, every seat, the operator's altitude.
Illustrative showcase · all names and numbers synthetic · design mirrored from live operating patterns · consulting & implementation: Vexera* · technology & operations: LOTL* · *working names