How it works

An administrator gives VERUS a system’s address and a read-only sign-in. VERUS lists what is there, reads it, and profiles every column. On the demonstration plant, a maintenance system’s three record types and 1,040 records were read in under a second.

VERUS works out what each table and column means. Where it is not sure it asks one short question at a time: “order type value Corrective looks like it means corrective. Is that right?” A person answers with one key. The maintenance system needed five answers.

Before anyone asks a question, VERUS checks for what makes the numbers wrong:

• a reason code used for more than half of all downtime, so totals by reason hide the cause;
• one shift recording scrap under the catch-all code and another never using it;
• 1,252 order lines whose promise date was moved to the ship date after shipping, so on-time delivery looks better than it was.

A plant manager’s dashboard of yesterday: build completion, labor efficiency, scrap and downtime, by line and department. It is run for that person, under their own access, and it says when its numbers carry a caveat.

“How much scrap did HB03 make on shift 2 in March 2026?” The answer is 93, far above any month in the year before. It shows how the question was read, the query that ran, and how fresh the data is. The model reads the question; the query is built by VERUS on the plant’s confirmed records, not written by a model.

An agent watches on-time delivery by customer. Last week four customers were below 85%. The plant manager accepted a plan for the first, with the measure it will be judged by: back to 85% within four weeks, from 71.4% when the plan was accepted.

“Warn me when scrap on any work center runs more than twice its usual in a month.” VERUS reads it back in plain words, and before it watches anything it shows what it would have raised over the past year: one finding, HB03 in March, $24,421 of scrap against a usual $10,705.

“What does the quality manual say about using scrap code 99?” The answer quotes the manual and cites the page. The March quality review of the same spike comes up beside it. Each person searches only the documents their role may read.

What the plant manager reads at six:

• a few lines on what matters most;
• yesterday’s numbers against the same day a week before;
• the findings that need a decision, and the plans under way;
• the outside world that touches the plant’s costs: steel mill product prices up 1.7% in August, plastic resins down 1.0%.

For a holding company, an investor or a board: every company at once, on one definition, read straight from the systems. Last month, Alpha Fabrication shipped $1.79M and Beta Extrusions C$735k, each in its own currency and never added together. On time to the original promise: 70.4% and 66.0%, with the warning that 1,252 order lines had their promise date moved after shipping. Scrap rate, downtime, headcount as a count, and twelve months of sales, a line for each company. A click on a company breaks it down by plant.

And an agent watching each company’s largest customers. Alpha’s largest, at $807,396 of orders over the past year, has had no call, visit or email since 14 June. Nobody told the owner; the CRM did.