Budget the answering layer against a missed call

Every plan that has crossed our desk this month has a line on it marked AI. Sometimes it is a number with no work attached, sometimes it is work with no number. Almost never is it sitting next to the thing it is supposed to replace, which is the cost of a phone nobody picked up.

That is the whole exercise for an answering layer. You are not buying a piece of software because it is the theme of the year. You are deciding whether covering the calls you currently lose is worth more than what covering them costs, and both halves of that are numbers already sitting in your own systems.

How we scope one of these, what goes in the brief and what an agent is never allowed to say are things we have written about before and are not going to run past you again. This is the budget page. What it costs once you count all of it, and what a year of listening to the recordings has changed about what we put in the brief in the first place.

The cost lines people leave out

A quote usually covers the first two of these. The budget needs all six.

• The platform. The subscription, the seat, the number. Small and predictable.
• Usage. Per minute or per call, which means it scales with your busiest month rather than your average one. Budget against January, or July if you are in home services, not against a quiet week.
• The build. The brief, the script written from your own recorded calls, calendar and CRM integration, the routing, and the test calls before it goes near a customer. One time, but real, and it is where a cheap setup shows.
• Weekly review. Somebody reads or listens to the transcripts every week for the first couple of months, then monthly. This is the line that gets cut in the negotiation and it is the line that decides whether the thing works.
• The downstream copy and process. The confirmation text, the intake fields, the handoff note to whoever calls back, the voicemail greeting that no longer makes sense.
• The change budget. You will rewrite the escalation rules in week two. Plan for that rather than being annoyed by it.

Nearly every deployment we have seen fail did not fail on the software. It failed because nobody owned the fourth line, so the agent kept making the same mistake for three months and the team quietly started routing around it.

What a year of transcripts changed in the brief

That fourth line is also the one that earns its money, so it is worth saying what came out of it. These are edits we now write into a brief before the first call rather than discovering in week three, because a year of recordings put them there.

1. Refusals got shorter. A long, apologetic refusal makes the caller explain themselves a second time. The line that works names the limit and offers the next step in the same breath, and it is one sentence long.
2. The callback number comes first, not last. Take it and read it back at the top of the call. A call that drops at forty seconds then still leaves you something to work with, and more of them drop than anybody expects.
3. No date of birth on a first call. Callers hear it as a gate and some of them hang up on it. Name, number, what they want. The rest happens with a person.
4. A synonym list built from the recordings, not from the website. People ask for a deep clean, a cap, a tune up, a quote for the whole thing. The words on your service pages are not the words people use on the phone, and the agent has to answer to both.
5. Two failed attempts is a handoff, and that is a counter rather than a judgment. We stopped trying to have it detect frustration. If it has asked for the same thing twice and still does not have it, it stops and fetches a person.
6. Hours questions are the largest category by a distance. Not bookings. So the hours answer gets its own written line, with the holiday exceptions in it, and it gets updated the same day the special hours change.
7. Nothing is added to the end of a call. No offer, no while I have you, no survey. It costs bookings, and it is the first thing a caller mentions when they complain about the whole idea.

None of that came out of a demo. It came out of somebody sitting down on a Friday with forty recordings and a notepad, which is the line in the budget everybody wants to delete.

The other half of next year’s budget belongs to the same idea from the other direction. The data you capture on those calls is first-party data, it is yours, and it is what feeds the follow-up when the ad platforms and the search answers give you less and less to look at. An answering layer that takes good notes is a data project wearing a phone headset.

If you resell this

One paragraph for the agencies, because we get asked and because there is an easy mistake sitting in it. Pass metered usage through at cost, with a stated monthly allowance and a named overage rate written into the agreement, so nobody is arguing about minutes at renewal. Take the margin on the supervision instead: the weekly transcript read, the brief edits that come out of it, and the escalation rules as a signed specification. Minutes are a commodity and any client can price check them in an afternoon. The Friday with the recordings is the part they cannot do themselves, and it is the part that decides whether the thing survives its first month.

What we are telling clients

• Put all six cost lines in the budget, not the subscription and the usage.
• Budget usage against your busiest month, not your average one.
• Name the person who reads transcripts every week and give them the hour in the diary, or do not start.
• Ask any vendor what their last three clients changed in the brief in month one. If the answer is nothing, nobody was reading the recordings.
• Decide which call types the agent may go near before you take a demo, because that decision is most of the price you will be quoted.
• Decide what you will stop doing to pay for it, which is usually the ad spend going to calls that ring out anyway.

Done in that order, this is an operations purchase with a number on both sides, which is the only kind worth signing in December. It is why AI answering sits in the same budget line as the CRM and the follow-up rather than in an innovation column, and it is how we scope it for the agencies we build for.