Ask an outage control centre what percentage of the critical path is complete and you will get a number. Ask where that number came from and you will get a person. Somewhere in the chain, a human being typed a status into a field, and everything downstream — the morning brief, the resequencing decision, the call to the plant manager — inherits whatever was true at the moment they typed it.
Nobody is careless. Updating a status is simply the third priority of a person whose first two are the job in front of them and the crew waiting on them. So the field decays quietly. P6 says four hours remain; the field finished the work yesterday afternoon. The schedule and the plant disagree, and neither one knows it.
Remove the place the drift lives
The alternative is not a better reminder, a stricter procedure, or an integration that syncs the field faster. It is to delete the field. In a system with no stored status, there is nothing to fall out of date, because state is not stored at all — it is computed on every read from evidence that carries its own timestamp and author.
A fact is asserted once: a worker signs, a Maximo work order closes, an Asset Suite attestation lands. Each carries a time, an author, and the decision record behind it. Release state is then an OR over AND-groups of required facts, recomputed at the moment of the question. Alternate paths need explicit human arming. Nothing is inferred and nothing is cached.
What the architecture buys you
- The picture cannot be stale, because there is no stored picture to age.
- Every blocker resolves to a specific missing fact, not a percentage.
- The missing fact has an owning role and a system of record, so escalation has an address.
- The critical-path hours behind that fact are computable, and therefore so are the dollars.
This is the difference between a dashboard that reports and a system that answers. A dashboard tells you the outage is 61 percent complete. An evidence-computed system tells you that the valve lineup is waiting on one signature, held by the RP supervisor, sitting in a permit that has not been closed — and that the wait is worth 6.5 hours of critical path.
The cost of the question you cannot ask
Plant outages are expensive… This can easily cost a plant operator more than $1 million a day. While the cost of planned outages is high, the cost of schedule overruns is even higher.
At that rate an hour is roughly $41,667. Any system that cannot name the fact holding the hour is, in effect, declining to price the wait. Radios, whiteboards and tribal knowledge fill the gap — and because the cost never appears on a screen, it never gets escalated to the person who could clear it in ten minutes.
None of this requires an agent to be clever. It requires the underlying model to be honest about what is known, when it became known, and who said so. The intelligence layer sits above that, and it is only as trustworthy as the layer beneath it.