Operations

Should it answer at all when nobody is there to take over?

Inside office hours an assistant is a filter in front of people who are there. Outside them it is the only thing there, and every limitation it has becomes the organisation's final answer until morning. That changes the calculation rather than extending it: some question types are safer unanswered than answered without a fallback, and the decision deserves to be made explicitly rather than inherited from the fact that software does not sleep.

Sort the inbound by what happens if the answer is wrong overnight

A question about opening hours, a policy document or a process step has the same correct answer at midnight as at noon, and getting it right unsupervised is fine. A question where being wrong causes action - a payment, a cancellation, a safety decision, anything time-critical - is a different category, because there is nobody to catch it before the person acts. Sort the real inbound into those two piles before deciding, using a month of actual messages rather than imagination.

Acknowledgement is worth more than most answers

A large share of out-of-hours contact is not seeking an answer so much as confirmation that the message arrived and will be dealt with. "Received at 02:14, the operations team reads this from 08:30 and you will have a reply before midday" closes most of them satisfactorily and commits nothing. This is the cheapest and most undervalued out-of-hours behaviour available, and it needs no knowledge base at all.

Decide what wakes somebody, and make that path real

If some categories genuinely need a person at three in the morning, the path has to exist before launch: whose phone, what threshold, what happens if they do not answer. Marking something urgent into a queue nobody reads until Monday is indistinguishable from not marking it. And if the honest answer is that nobody is available, the assistant should say so plainly along with what the person can do meanwhile - which beats an implied promise.

The overnight transcript is the best diagnostic the organisation gets

What people ask when no one is watching, in their own words, with nobody to rephrase it for them, is a cleaner picture of what the organisation is unclear about than any survey. Read a week of it before deciding whether to widen the overnight scope - and expect the list of what to document to be the more valuable output.

Frequently asked questions

Is running an assistant after hours different from during them?
Yes, because the escalation path is missing. Inside hours it filters in front of people who are there; outside them it is the only thing there, and every limitation becomes the organisation's final answer until morning.
Which questions are safe to answer unsupervised?
Those with the same correct answer at midnight as at noon - opening hours, a policy, a process step. Questions where being wrong causes action - a payment, a cancellation, anything time-critical - are a different category, because nobody is there to catch it.
What if we can only do acknowledgement?
That is already worth most of the value. "Received at 02:14, the team reads this from 08:30, you will have a reply before midday" closes most out-of-hours contact satisfactorily, commits nothing, and needs no knowledge base.
Do we need an escalation path at three in the morning?
Only if some category genuinely requires one - and then it must exist before launch, with a named phone and a threshold. An urgent marker in a queue nobody opens until Monday is the same as no marker.
What should we read before deciding?
A month of actual out-of-hours messages, sorted by what happens if the answer is wrong. The overnight transcript is a cleaner picture of what the organisation is unclear about than any survey, and the list of what to document is usually the more valuable output.

Get in touch

Leave your details and we will reply. You do not need a sales call to get an answer.