The Most Expensive Door in the Building

Every support organization has a price list it has never written down. It is the cost of the same customer question arriving through different doors, and the spread between those doors is far wider than most teams assume.

Per Unthread's cost-per-resolution research, phone runs $17 to $25 per resolution. Human live chat lands at $10 to $16. Email sits at $8 to $15. Self-service costs $1 to $4. Gartner's figures tell the same story from another angle: $1.84 per contact for self-service against $13.50 for assisted channels, roughly a sevenfold gap, with AI chatbots near $0.50 per interaction versus about $6.00 for human-handled chat.

Same question. Same customer. Five to twenty times the cost depending on which door they walked through.

Deflection strategy usually starts in the wrong place

Watch how most organizations sequence an automation program and a pattern shows up immediately. Chat goes first.

The reasoning is understandable. Chat is text, text is easy for a model, the widget is already on the site, and nothing about it feels risky. So the pilot launches on the cheapest channel in the building.

It works. It deflects. And it moves almost nothing, because the money was never in chat.

Sequencing by cost rather than by ease inverts the roadmap. Phone is the most expensive contact an organization can accept, so every deflected call is worth several deflected chats. A program that removes 20% of phone volume beats one that removes 50% of chat volume in most cost structures, and it usually is not close.

Why voice gets postponed anyway

Two reasons, one good and one bad.

The good reason is that voice is genuinely harder. Latency is unforgiving, interruptions are constant, and a bad transcription in a live conversation is far more damaging than a bad autocomplete in a chat window. Teams that have waited for the tooling to mature have often been right to wait.

The bad reason is that phone volume is invisible in exactly the way that protects it. Chat and email leave a searchable record. Calls leave a duration and a disposition code somebody selected in a hurry. It is difficult to argue for automating a channel nobody can characterize, so the channel keeps its budget by being unmeasured.

Any organization serious about cost per contact has to break that first. Not with a vendor. With a transcript sample large enough to answer one question: what share of these calls are people asking something the system already knows.

The deflection numbers are real, and they are not the whole story

The published results support the effort. Unthread reports AI deflection exceeding 45% of incoming queries and a 55% reduction in average first-response time. Lemonade reached 40% automatic resolution across IT, HR, Legal, Procurement, and Finance with purpose-built agents rather than one general assistant.

That last detail is the one worth stealing. The gains came from several narrow agents matched to specific workloads, not from a single system asked to be good at everything. Narrow scope is what makes a deflection rate defensible, because a purpose-built agent has a knowable failure mode and a general one does not.

The sequencing argument, plainly

An automation roadmap ordered by technical comfort produces a portfolio of small wins in cheap channels and a phone queue that costs exactly what it did before.

An automation roadmap ordered by cost per contact starts in the uncomfortable place, moves slower, and shows up in the P&L.

For property management, healthcare, and any operation where the phone is still the default channel for anything urgent, this is not a close call. The steepest per-contact savings sit precisely where the work is hardest.

The cheapest channel to automate and the most expensive channel to keep are rarely the same channel. Most roadmaps quietly optimize for the first one.

Sources: Unthread (phone $17-$25, email $8-$15, human live chat $10-$16, self-service $1-$4 per resolution; AI deflection exceeding 45% of queries; 55% reduction in average first-response time; Lemonade 40% automatic resolution across IT, HR, Legal, Procurement and Finance); Gartner via Unthread (self-service $1.84 per contact vs $13.50 assisted; AI chatbot ~$0.50 per interaction vs ~$6.00 human-handled chat).