CASE STUDY — Decision OS · DECISION INTELLIGENCE · OPERATIONS ANALYTICS

Decision OS: $5.8M Caught Before It Landed.

A team with dashboards still files tickets, because the question that decides something is the one nobody built a dashboard for, and that answer comes back in three to five days. The hard part was never writing the query. It was that by the time the answer arrived, the decision had already been made without it.

BY Jen, AI & Data Lead | September 8, 2026 · SEPTEMBER 8, 2026

$5.8M

CAUGHT BEFORE IT LANDED, ACROSS REFERENCE DEPLOYMENTS

$5.8M CAUGHT BEFORE IT LANDED, ACROSS REFERENCE DEPLOYMENTS
UNDER 6 SECONDS TO AN ANSWER — AGAINST AN ANALYST QUEUE OF 3–5 DAYS
9.4 DAYS MEDIAN WARNING BEFORE THE EVENT
83% OF ALERTS CONFIRMED BY OUTCOME · 41 FIRED LAST QUARTER

01 — THE PROBLEM

Three to five days, and four bad ways to shorten it

Three to five days to an answer. One question per ticket. A follow-up is a second ticket. The decision, meanwhile, gets made. The question that decides something rarely arrives on a schedule. It surfaces mid-meeting, from a COO asking which stores are bleeding margin this week. It joins the queue. Four ways to shorten that queue. Hire analysts. The queue scales with them. Build more dashboards, and you have only widened the set of questions someone thought of in advance. A query box tells a retail operator that markdown is running hot, then leaves them to find where. Self-serve SQL is worse: most operators stop asking. The question does not get answered later. It stops being a question. A follow-up costs another ticket and another three days. Decision OS answers it in the same conversation.

A queue does not just delay the answer. It decides which questions are worth asking.

02 — HOW IT ANSWERS

Four layers, because a bare number restarts the argument

The six seconds is the least interesting thing about Decision OS. A question goes in as plain English; the system writes the query, runs it against the customer's own data and composes the reply. Under six seconds, measured against a human process running in days rather than against a latency spec. The shape of the answer matters more. A bare number restarts the argument: the next thing anyone asks is where the overrun is coming from. So the reply carries four layers. Take the markdown example the console runs. $1.9M of unplanned markdown, 6.9% of sales this quarter against a 4.2% plan. The leak: 61% of that overrun sits in 74 stores clearing above the 22% threshold without sign-off. The slope, charted in the reply, $3.1M by Q4 if nothing changes. Then three ordered steps, stores named, ready for the category team that afternoon.

03 — MAKING THE ANSWER TRUSTABLE

It runs in the customer's cloud, changes nothing, and shows its working

An answer in six seconds is worth nothing if nobody downstream believes it, so the constraints count for more than the features. Decision OS runs inside the customer's own cloud account, next to their data. It reads; it never writes a record. Every person sees only what their role already allows. Nothing asked of it trains anyone's model. Believing a number takes more than that. Every answer carries the exact query that produced it, so any figure can be taken apart by whoever defends it, and a panel headed what's weak about this number sits in the same reply. The bar is a regulated healthcare group running it on live patient records. It runs there now.

04 — WHAT IT BECAME

Nine days of warning instead of a line in the monthly report

Trust showed up as a change in timing. Once operators believed the answers, they asked before the month closed, not after. Every metric gets a forward curve fitted to it, so an alert fires where the curve crosses a threshold. Median warning before the event: 9.4 days. Of the alerts it fires, 83% are confirmed by outcome. Forty-one fired last quarter. In the published worked example, warehouse 04's stock cover was falling fast enough to cross the reorder threshold in nine days. The alert fired at day minus nine; the expedite window stayed open until day minus two, so seven of those days were usable. By day zero, where a monthly report would first have mentioned it, $610K was gone. A live alert reads the same way: apparel inventory crossing 74 days before the peak-season order has to be placed, $740K a quarter in carry. What we will not claim: there is no per-customer breakdown of that $5.8M, and the deployments behind it are reference deployments, not a named list. Where a first caught leak has covered the year, that is what those deployments showed. The test we watch is duller. Operators stopped filing the ticket.

Tired of waiting
on the ticket queue?

Talk to the team Agentic AI service →

FAQ

Questions about this product

We just rebuilt our BI stack. How is this not another dashboard?

It isn't one, and it doesn't replace one. Your dashboards will keep answering the questions they were built for, instantly. The thing to compare it against is the ticket. In the queue a question gets answered as written, once, and the follow-up is a second ticket and another three to five days. Here the follow-up costs nothing. You ask again, in the same conversation, until you have enough to decide.

What happens when it is confidently wrong?

You can check it, and you are told where to look. Each answer ships with the generated SQL and its audit trail, and each one carries a panel headed what's weak about this number. On the forecasting side, 83% of alerts are confirmed by outcome. That is a published rate, and it also means roughly one alert in six does not play out. On anything that moves money, someone still reads the query first.

Our data cannot leave our environment. Does that rule us out?

No, and you can test that before you sign anything: the demo runs against your own schema, not ours. In production it sits inside your own cloud account, read-only and role-aware, and nothing leaves that account, the generated SQL included. So the record of what was asked and what came back stays with you. It already runs in a regulated healthcare group on live patient records.

How long before anyone actually gets value from it?

Week one is a read-only connection and your first ten real questions answered live, in the room. Weeks two to four, operators ask directly and the queue drains, which is also when your analysts stop working as a service desk. Month two, the alerts run against forward curves on margin, stock and working capital. No migration, no new dashboards to build, no retraining. On quarter one we say only this: the first caught leak has typically covered the year on the deployments we have.