This describes work that is in design. Everything below is what we are building, not something you can switch on today.
Friday, 5:55pm. An inquiry lands about the downtown building.
It goes to leasing@, which four people can see. Each of them assumes somebody else is watching it over the weekend, which is a reasonable assumption for all four to make and produces the outcome you would expect. The person who actually owns that building is off until Tuesday. The person covering the weekend does not have the keys and has never been inside the unit.
The applicant books nothing. By Monday they have seen three other places.
Nobody made a mistake. Not one person did anything wrong, and if you ran an inquest you would find four people who each behaved sensibly. The inbox is the mistake.
The failure that costs the most is a dropped one, not a wrong one
Every firm worries about giving an applicant the wrong answer. Very few worry about giving them none, and the second is far more expensive.
A wrong answer at least generates a conversation. Somebody replies, you correct it, and the applicant is still in the process. A dropped inquiry produces nothing at all — no complaint, no correction, no record. The applicant simply proceeds with their life somewhere else, and your firm's experience of the event is that nothing happened.
This is the same thing article five in this series described as a handoff count: things break where work passes between people, and the very first handoff — inquiry to whoever answers it — is the one that fails first, because it is the only stage that leaves no artefact behind when it fails.
leasing@ is a handoff with nobody on the far side of it.
The facts your firm already has
Here is what makes this tractable, and it is the whole design argument.
Your firm already knows who owns which buildings. It knows who covers weekends. It knows that anything about a lease term goes to Marie, that the heritage building is the owner's personal project, and that the person handling the north end is off on Thursdays.
None of that is written down anywhere the software can see. It is tribal knowledge — real, stable, universally understood inside the firm, and completely unavailable to any system that might act on it.
So the design decision is unglamorous: make ownership and coverage first-class configuration rather than institutional memory. Who owns which properties. Who covers which periods. What categories of question go to which person regardless of property. Written down once, in the same place as everything else the firm has configured.
Once that exists, a conversation reaches a person rather than a shared address, and the weekend inquiry about the downtown building goes to whoever is actually covering the downtown building this weekend — not because the software worked out who was probably free, but because somebody wrote it down.
The other half: slots that are genuinely in somebody's day
The same facts fix a second problem, which is subtler and which applicants feel directly.
When a calendar offers times without knowing who would actually be standing there, a confirmed showing is a hope. The slot exists in the abstract — the property is free at 2pm Thursday — and whether any human being is available to attend it is discovered later, usually by the human being.
What we are designing is that availability offered to an applicant comes from real availability: who is free, who is off, who is covering that building. So a booking that gets confirmed is genuinely in somebody's day.
One thing about how that has to work. Availability is computed by deterministic code, and the software's job is to offer exactly what the engine returned. It does not reason about who is probably free, or who ought to take this one, or who has had a light week. Scheduling that runs on inference produces a calendar nobody trusts, and a calendar nobody trusts gets checked manually, which returns you to the situation you were trying to leave.
Routing is about your colleagues, never about applicants
One boundary, stated plainly, because the vocabulary of this feature area drifts toward people if you let it.
Routing decides which colleague handles a thread. It never orders, ranks or selects among applicants. There is no priority queue of people, no triage of who is worth answering first, no sorting of a morning's inquiries by anything about the person who sent them.
Everybody who writes to your firm gets routed to whoever should handle it. The routing is about your team's structure, and the applicant's place in the queue is when they arrived.
The trade-off: a coverage map goes stale
Now the honest cost, and it is the same failure that afflicts any written-down thing.
Routing is only as good as the coverage map. And coverage maps rot faster than policies do — a person leaves, territories get reshuffled after a portfolio acquisition, somebody swaps weekends for a month and it never gets swapped back. Meanwhile the rule keeps routing confidently to a person who no longer covers that building, and it does so silently, which is worse than not routing at all.
So routing rules get the same treatment as policy lines: a named owner and a review date. Somebody is responsible for the coverage map being true, and the map is reviewed on a schedule rather than when it fails. A rule that has not been confirmed in a long time should not be trusted to send a Friday-evening inquiry to somebody's dormant inbox.
The general principle underneath both: anything a firm writes down for software to act on needs an owner, a date and a way to see when it was last true. That applies to the pet policy and it applies to who covers the weekend.
What Friday at 5:55pm looks like instead
Same inquiry, same building, same four people, same Friday.
It reaches the person covering the downtown building this weekend, because that fact is written down. They know it is theirs — not because they happened to look at a shared inbox, but because it arrived addressed to them. If the criteria are met and there is real availability in somebody's Saturday, the applicant books it themselves.
And if it turns out to be a question that should go to Marie, it goes to Marie, because that is also written down.
Nobody had to be watching an inbox on a Friday evening. That is the whole ambition here: not that the software is clever about your team, but that your team's actual shape stops being something the software cannot see.