Voice out-of-scope: connect the call, or take the message?

• Clara never opens with what she can't do. Shipped 2026-08-14 on every channel. The caller hears the next step first.
• When a caller asks for a person, they get one. Any line, any wording — the call transfers immediately. Same for a vendor standing on-site who needs access now.
• Wording ruling (Fede, 2026-08-15): when Clara does connect a caller to a human, she says "Let me connect you with someone on the team." This is a small copy change (today she just says "one moment") and is safe to ship on its own — announcing a transfer is only banned between Clara's own invisible specialist handoffs, not when a real human picks up. It ships as its own small PR on your go, whatever you decide below.

1 · The office phone starts ringing more, including for junk

Today the unknown-caller line quietly absorbs solicitors and wrong numbers as low-urgency notes. Forward-by-default sends those to the office too, unless we carve them out — and a carve-out means Clara deciding live who "deserves" a human, which is its own failure mode.

2 · After hours, the caller's outcome gets worse, not better

You ruled no-answer → office voicemail. But today's after-hours caller gets a structured note in the team inbox with urgency and callback number attached; a voicemail is unstructured, easier to miss, and the note never gets created. The caller also sits through ringing before learning nobody's there.

3 · It trades away part of what Clara is for

The pitch to the property team is that Clara absorbs interruptions so humans work uninterrupted. Forward-by-default moves calls back onto the office line — the metric moves in the direction the product exists to prevent. Worth doing only if the caller-experience win clearly outweighs it.

4 · We don't know the volume yet

No number exists today for how many calls per week this default would redirect (the count of messages taken where the caller never asked for a person). That's measurable from the last 30 days of unknown-caller notes and transfer outcomes before committing.

All voice calls from the last 30 days were pulled from the call platform: 588 total, of which 150 were real-property calls (Camellia + Yale 25) — the rest were our own test robots and bench agents. Classing office hours as Mon–Fri 9–5 Denver (assumption — exact posted hours may differ slightly):

• 117 calls during office hours, 33 after hours — about one after-hours call a day.
• Of the 33 after-hours calls: 6 already ended in a live transfer (caller asked for a person — all connected), 1 became a captured message, 1 paged the manager, and 25 needed neither: silent or abandoned calls, quick hang-ups, a tour that got booked, a maintenance issue that routed correctly, and several calls Clara herself placed (renewal callbacks, a vendor ETA).
• The calls forward-by-default would actually redirect — a caller with a real out-of-scope need who today gets a message instead of a ring — came to roughly 1–2 per week, and some of the counted traffic is likely our own test callers dialing the live line, so the true number is if anything lower.

Every figure above traces to the pulled call records; nothing is estimated except the office-hours boundary, which is labeled.

A · Measure first — done, see the section above

Fede picked this on 2026-08-15; the numbers are in. The remaining choice is B, C, or D below.

C · Forward by default on all six lines

Consistent everywhere; largest shift of live calls to the office. Not recommended without the measurement.

D · Keep today's behavior, ship only the wording ruling

Messages stay the default; the already-settled improvements (handoff-first openings, transfer-on-request, "let me connect you" copy when transferring) still land. Zero new load on the office.