Question 1

How do you recognize it?

Accepted Answer

Client call logs show 40 to 60 percent of connected calls running past four minutes, and the per-minute invoice lands above the flat monthly fee the agency quoted at signing. The agency's delivery lead is manually editing call transcripts and re-recording prompt corrections every week, work that was never scoped into the retainer. Two consecutive invoices show usage cost climbing while the client's booked-appointment count stays flat, so the client questions what the fee is buying. Escalation to a human happens on fewer than one in five calls, but the ones that do escalate consume 15 to 20 minutes of a senior staffer's day. The client's phone provider bill and the voice platform bill arrive as separate line items, and nobody on the agency side reconciles them before the client does.

Question 2

Why does it happen?

Accepted Answer

Agencies price the retainer off the demo call, which runs two minutes with a cooperative caller, then discover production calls include hold time, repeats, and voicemail detection loops that multiply billed minutes. Usage-based components (speech-to-text, text-to-speech, telephony, and model inference) are metered separately by most platforms, so a single quoted per-minute rate hides three or four stacked costs. Consent and recording disclosure requirements in some jurisdictions force a longer opening script, adding 15 to 30 seconds to every call before any business value is delivered. The labor that remains after deployment, including transcript review, prompt tuning, and escalation handling, is treated as launch support rather than an ongoing delivery cost.

Question 3

How do you fix it?

Accepted Answer

Pull 30 real call recordings from the client's first two weeks, calculate the true average call duration including abandoned and voicemail calls, and rebuild the retainer math on that number before the next invoice goes out. Separate the client-facing fee into a platform line and a delivery line so usage spikes are visible to the client rather than absorbed silently by the agency. Set a hard escalation rule with the client (for example, transfer to a human after two failed intent matches) and price the human handling time into the monthly fee. Cap monthly included minutes in the contract and define an overage rate, so a busy month does not convert a fixed retainer into a loss.

More for AI Voice Agent