Operations

Operations is where the cost of repetitive work is hardest to see, because it is spread thin. Nobody spends a day on status updates. Everybody spends twenty minutes on them, several times a week, and the total is invisible until someone counts it.

The second cost is worse and even less visible: the work that does not get done because the update had to be written.

Scheduling changes are a cascade, not an event

One job moves. Three other things depend on it, two people need to know, and one customer has been given a time that is no longer true. The rescheduling itself takes a minute; the notifying takes twenty and is the part most likely to be skipped under pressure.

An agent handles the cascade. It knows what depends on what, it works out who is affected, and it tells them. What it does not do is decide which of two colliding jobs takes priority, because that decision usually depends on a relationship, a promise or a risk that exists nowhere in the schedule.

Status updates written from what happened

Most status reporting is a person reconstructing a week from memory and a calendar, then rewriting it for whoever is reading. The reconstruction is the expensive part and the least reliable.

An agent assembles the update from the record: what moved, what completed, what is blocked and for how long. That draft is more accurate than a reconstruction and takes no time to produce. What it lacks is emphasis, which is precisely what a person adds: the three lines that say what actually matters this week.

That division tends to hold across all of this work. The agent produces the complete version; the person decides what to point at.

The handover is where operations actually leaks

Work passing between people is the point at which context is most often lost, and the loss is silent. The next person starts from what was written down, which is rarely everything, and the gap surfaces days later as a mistake nobody can trace.

A handover assembled by an agent carries the full record rather than the summary someone had time to write at five o’clock. It is less elegant than a well-written note and considerably more complete, and completeness is what handovers fail on.

Where this stops being useful

Operations that change shape every week are not a good candidate. Rules that shift constantly cost more to maintain than the work they replace, and each change means revisiting the boundary rather than tweaking a setting.

If your process is still being invented, automating it locks in a version you are about to outgrow. Write it down first; the document alone often removes more friction than the software would have.

How this gets built

Mapping comes first, and in operations that map is usually the deliverable people value most, because the process has never existed on paper. Then one flow goes live, usually status updates, since a wrong update is visible immediately and costs nothing to correct. The complete sequence is under how a project runs, and you can stop after the mapping stage with the document in hand.

Operations tends to touch several systems at once, so the scope is worked out during the mapping rather than estimated beforehand. Book a demo to see what that mapping produces.

The judgement about which of these flows is worth moving across is the subject of where AI agents actually save time.