Natural language supply chain analytics — ask in plain words, see the query behind every number
Ask any supply question in plain words and get a traced answer in seconds.
See one case, screen by screen ↓Backorders sit in order management, stock in the warehouse system, shipments in transport management and the forecast in demand planning. Each has its own report. “Which backordered SKUs have stock arriving this week?” needs two of them joined on SKU, with the same definition of “EMEA” and “this week” on both sides.
A fixed dashboard answers the questions it was built for. A new one goes to an analyst as a request and waits in the queue — and by the time the numbers come back, the shipment has landed or the hospital has called. The planner who needed the answer has already decided without it.
Estimated split for one cross-cutting question raised as an analyst request, and asked directly instead, in working hours.
Six agents screen, route and answer every question read-only; a supply chain leader validates answers.
Monday, Oct 5 to Sunday, Oct 11, 2026. Lena Ortiz runs supply for EMEA and wants to know which backorders this week’s arrivals will clear. Here is what happens, screen by screen, in the working solution.
Lena opens the solution: 214 questions asked this week by 31 people, a median of 3.4 s to an answer, 2 questions stopped by the guardrails. Under “Needs you”: a forecast answer waiting for her validation, 7 SKUs below safety stock at Plant 3 — Leiden (up from 4 yesterday), and a cold-chain shipment that logged 8.6 °C for 22 minutes.
Watched numbers: “Nightly validation — 34 real questions · 34 / 34 · all passing.”
The question is screened, and the orchestrator sends “backordered” to the Backorders specialist and “in transit arriving this week” to the In-transit specialist. The answer: 6 of the 14 SKUs on backorder in EMEA have stock arriving — 7,800 units against 6,295 backordered. Four are fully covered; CVT-SH-7F11 and NXL-DES-2515 stay short by 490 units between them.
“NXL-BC-3020 lands Oct 10 but needs incoming quality release, so it can ship from Oct 12.”
Under Query & reasoning sit the two SELECT statements: the Backorders specialist read 1,284 open lines in 0.8 s, the In-transit specialist 37 shipment lines in 0.7 s. The reasoning spells out every assumption — “EMEA” is the order’s ship-to region, not where the stock sits; “this week” is Oct 5 to Oct 11 by carrier ETA; stock already received counts as on hand, not in transit.
“The orchestrator joined the two results on SKU in the answer itself — the warehouse only ever saw two read-only queries.”
Lineage lists the four tables read: backorder lines from order management (ERP), refreshed 06:10; the product master, 02:00; inbound shipments and shipment lines from transport management, 06:05 — with the row count read from each.
“Every number in this answer comes from these tables. Nothing was written back.”
The follow-up carries over the 6 SKUs from the last answer. The Forecast specialist reads the approved September S&OP version for the next four weeks; the Inventory specialist reads free stock and confirmed receipts. Two SKUs stay at high risk: CVT-SH-7F11’s shortfall grows from 280 to about 1,480 units before its Nov 2 receipt, and NXL-DES-3018 runs out around Oct 17, about 300 short until Oct 21.
Bias shown so the forecast can be judged: CVT-SH-7F11 demand ran 14% above forecast over 8 weeks.
At 08:03 Priya Raman, planning APAC, typed “Update the ETA for SHP-EU-40851”. The question screener reads it as a request to change data and stops it — asked again here, it gets the same stop: no SQL written, 0 tables touched, the full text logged for the quality reviewer.
“Supply Chain Answers only reads. Shipment ETAs, orders and stock are changed in their own systems by the teams that own them — nothing was sent to the warehouse.”
Lena saved the forecast answer from Ask at 06:14. In the Library it sits beside the team’s other saved answers — re-run after the 06:10 warehouse load, scheduled or not, each marked Validated, To validate, Numbers changed, Quality review, Shared or Up to date. This one says To validate.
Validating asks her to confirm three things: she checked the numbers against the source tables, the reasoning matches how the team defines these terms, and the expected answer is saved for the nightly runs — match rule: totals within ±1%.
“A validated answer joins the shared set. If the answer drifts, you are asked to review it.”
The history now reads: Lena Ortiz validated it — signed, expected answer saved, asked again tonight at 02:00. The nightly set grows from 34 to 35 questions.
“Validated — now 35 questions in the nightly set.”
At 02:00 every validated question is asked again and compared with its expected answer. The dashboard shows each night’s score — one night in the last seven came in at 33 of 34 — beside questions answered per week, which specialists answered, time to an answer, and every question the guardrails stopped.
The same warehouse, seen by the five people who rely on it — what their week looked like, and what it looks like now.
Six agents screen, route and answer every question read-only; a supply chain leader validates answers.
Checks every question before it is routed — stops requests to change data, requests for personal or restricted data, and instructions hidden in a question.
Reads each question, picks the specialists, chains them for multi-part questions, passes the SKUs, regions and weeks between them, and joins their results into one answer with a table and a chart.
Answers questions about open backorders — which SKUs, which customers, how old, how much.
Answers questions about free stock, safety stock, shelf life and confirmed receipts by site.
Answers questions about inbound shipments — what is moving, when it lands, delays and cold-chain status.
Answers questions about demand forecast, accuracy and bias by SKU, family and region.
Validates the answers the team shares. The agents propose; a named person decides.
Beyond the questions themselves, the assistant can explain an answer, set a watch on a number, schedule a saved answer for a colleague or request a table — always inside the same read-only limits.
Which EMEA SKUs are still short after this week’s arrivals?
Two stay short: CVT-SH-7F11 by 280 units (next receipt Nov 2) and NXL-DES-2515 by 210 units (clears with the Oct 14 receipt). CVT-SH-7F11 is the one to watch — demand has run 14% above forecast for eight weeks.
Alert me when CVT-SH-7F11 backorders pass 1,000 units
Done. A watch on CVT-SH-7F11 backordered units in EMEA — 880 today. The Backorders specialist checks it every hour after the warehouse refresh and notifies you when it passes 1,000. It reads the same table as your saved answer, so no new access was needed.
Send the EMEA arrivals answer to Sam Patel every Monday at 7:00
Scheduled. Every Monday at 07:00 the answer re-runs against the 06:10 warehouse load and goes to Sam Patel with the SQL attached. If a number moves more than 10% since the week before, it is marked Numbers changed so you see it first.
Let the Forecast specialist read supplier lead times
A request is raised to add md.supplier to the Forecast specialist’s tables. It already serves Inventory and In-transit. Omar Haddad owns the data and decides; until he approves, the Forecast specialist cannot read it.
13 screens from the working solution, on its sample data. Pick one to see it large.
Which specialists run, how far a query may go, who approves and how often answers refresh are settings.
| Setting | Default | Choose from |
|---|---|---|
| Read-only queries | Always on | Cannot be turned off |
| Rows per query | 5,000 | 1,000 to 10,000, in steps of 1,000; larger results are cut and the answer says so |
| Time per query | 30 seconds | 15 · 30 · 60 seconds |
| Screen every question | On | On · Off |
| Add a table to a specialist | Omar Haddad | Any named person |
| Validate an answer for the shared set | Lena Ortiz | Any named person |
| Refresh saved answers | Every morning, 06:15 | Every morning, 06:15 · Every hour · Only when opened |
| Flag numbers that move | 10% | 5 · 10 · 20% |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample supply-chain data · “target” = the design goal, measured in the live solution · “estimated” = our estimate. People, companies and products named on this page are fictional — characters and sample data in the working solution.
Asking a supply question in plain words — “which SKUs at Plant 3 are below safety stock?” — and getting the answer straight from the source tables, with the query shown. Supply Chain Answers routes each question to specialists for backorders, inventory, in-transit and forecast, and returns a table, a chart and the SQL behind every number.
No. Only SELECT statements reach the warehouse, enforced on the connection and not just by the agents. Requests to change data, such as updating an ETA, are stopped by the question screener before any query is written.
Every answer shows the SQL each specialist ran, the assumptions it made about dates, regions and statuses, and the tables it read with their refresh times. Answers the team shares are validated by a named leader and asked again every night against their expected result.
Yes. The orchestrator chains specialists for multi-part questions and joins their results on shared keys such as SKU. A follow-up carries over the SKUs, regions and weeks from the previous answer, so nothing is asked twice.
The data owner. Each specialist reads only its own allow-listed tables, supplier prices are restricted for all of them, and a new table needs the data owner’s approval before any specialist can read it.
Yes. The specialists only read and answer. A named leader validates answers before they join the shared set, a person reviews any nightly drift, and every question, query and stop is in the audit trail.
The Agentic Solution Engine builds and deploys it from your requirements — your table catalog, definitions and a set of real questions with agreed answers — and it goes live once every quality gate has passed. We will walk you through it on your own supply questions first.
An analyst builds a new data specialist by conversation in minutes, with no code.
1-min filmSupply Early WarningEvery supply signal read as it lands, and a ranked, cited shortage alert raised weeks before stock runs out.
1-min filmTemperature ExcursionsEvery temperature alarm becomes a cited release-or-reject decision in hours, with every earlier leg counted.
1-min filmWe’ll run Supply Chain Answers on a sample of your own backorder, stock and shipment questions.