Resolution Agent

Genuine self-service

A customer resolves their own issue without opening a ticket. The tour follows the whole path — including the point where the system decides it cannot help and routes to a person, rather than looping the customer through dead ends until they give up.

  • The customer resolves it without opening a ticket
  • Grounded in your documentation and resolved cases
  • Stops and routes to a person when it should
  • Deflection only counts if the answer arrived

Watch the full tour

Verify your work email once and every research paper, case study and product tour on ascendo.ai opens, right here, without leaving the page.

  • Takes about 30 seconds
  • Work email only, no spam

Deflection and resolution are not the same number

Every self-service programme reports a deflection rate, and most are measuring the wrong thing. A contact that did not become a ticket is not necessarily a customer who got their answer. It might be a customer who gave up.

Those two look identical in the reporting and are opposites in reality. One saved you a ticket; the other cost you trust and comes back next week through a more expensive channel.

Four different customer outcomes — helped, gave up, went looking for a phone number, and returned angrier a week later — all passing into a single deflection count that reports them identically.
Three of these four outcomes cost you something, and the metric cannot tell any of them apart from the first.

A path that actually ends somewhere

The fourth step is the one that makes the first three trustworthy.

  1. Step 1

    Understand the problem

    What the customer describes is read as a problem rather than matched against article titles.

  2. Step 2

    Answer from real material

    The response comes from your documentation and the cases your team has resolved, which is what makes it specific enough to fix something.

  3. Step 3

    Work through it with them

    The customer is taken through the steps rather than handed a document, with what they report back shaping what comes next.

  4. Step 4

    Stop when it is not working

    Where the path is not leading anywhere it routes to a person with everything gathered attached.

Four stages: the customer’s description read as a problem, an answer drawn from documentation and resolved cases, the customer taken through the steps, and a route to a person when it is not working.
The fourth stage is what makes the first three trustworthy — and it is the one that costs a deflection to build.

What separates this from a help centre

One of these is a design decision most vendors avoid making.

The exit is what makes it honest

Any self-service system can keep a customer busy. What separates a good one is knowing when it is not helping and saying so early. That costs a deflection and earns a customer.

It answers from your material, not article titles

Traditional self-service matches the customer’s words against a document library. Answering from your own material is the difference between being handed something to read and being told what to do.

Measure resolution, not avoidance

A deflection rate counts contacts that did not happen, which includes every customer who gave up. What is worth measuring is whether the problem went away.

A five-row table of self-service outcomes showing which count as a deflection, which actually resolved the customer’s problem, and what each one costs afterwards.
The second column is the one worth collecting, and the third is what the first column is hiding when it looks healthy.

Why the usual approaches fall short

Three of them report well and help less than the number suggests.

A help centre with search

Matches words against article titles and hands over a document.

A chatbot over an FAQ

Conversational packaging on the same lookup, and it usually fails by looping.

Deflecting on a keyword

Sends an article and closes the contact. It counts as a deflection whether or not it helped.

No self-service at all

Every routine question reaches a person, including ones a customer would have solved at midnight.

Two columns comparing a help centre with search, a chatbot over an FAQ, keyword deflection and having no self-service with a guided path that routes to a person when it stalls.
Keyword deflection reports best of all four, because it counts the contact as saved whether or not the article helped.

What things are called

Deflection and resolution are the pair worth keeping apart.

Deflection
A contact that did not become a ticket. Not the same thing as a resolution.
Containment
The share of conversations handled without a person. Only meaningful alongside whether the problem was solved.
Escalation path
The route to a human when self-service is not working, and what carries across when it fires.
Grounding
Restricting answers to your own documentation and resolved cases.
AI Resolve
The pluggable Gen AI component that can be dropped into any website or portal.

Frequently Asked Questions

By what happened afterwards rather than by the deflection itself: whether the customer returned, whether the same issue arrived through another channel, and whether the conversation ended with an answer or with silence. A deflection rate on its own tells you none of that.

Where this fits

Bot to human handoff

What carries across at the moment self-service stops and a person takes over.

Multiple AI bots tour

Running a different assistant per product line, segment or channel over one knowledge foundation.

AI Resolve

The pluggable component for putting this into an existing website or portal.

More Resolution Agent tours

All tours

Support Channels

Check what happened after your deflections

Take last month’s deflected contacts and look at who came back. We will show you the difference between avoided and resolved.

Talk to us