STEN.AI

Proposed test Customer service

Can a company resolve service requests with less rework?

Many AI tools for customer service promise faster answers. What counts is whether a case stays solved after quality checks and whether people spend less time putting answers right.

Why this question

Buyers can rarely compare offers on facts.

ETHORITY's market study reviewed what 107 customer service entries disclose on their own public pages. Fewer than half show a price, and prices follow at least seven different models. To compare, a buyer has to calculate each offer at the same expected case volume and confirm the hosting region and data terms in writing.

43of 107 entries show a public price
38of 107 entries document an EU hosting option

Prices follow at least seven different models, including per resolved case, per seat and per minute, as well as quote-only offers.

Source: Good AI Prism Market Study 2026/27, Global edition v2.6, 27 Oct 2026, pp. 54–55. Vendor information checked 19–20 Oct 2026. The 107 are entries in the research set, not a count of distinct products.

The test

Six steps, agreed before the start.

  1. 01

    Define what should improve

    Cases that stay solved, less time spent on rework and fewer repeat contacts. The customer's view counts as much as the team's.

  2. 02

    Record the baseline

    Measure the current state over an agreed period: reopen rate, handling time including quality checks and rework, repeat contacts and handovers to a person.

  3. 03

    Limit the deployment

    One request type, one team. The AI assistant drafts answers, and a person checks every answer before it goes out.

  4. 04

    Compare offers on the same basis

    Calculate each offer at the same expected case volume and confirm the hosting region and customer data terms in writing.

  5. 05

    Agree the threshold and the decision owner

    Before the start, the head of service and the team agree which result justifies expansion. One named person decides.

  6. 06

    Decide

    Expand to the next request type, change the set-up or stop. The decision and its numbers are documented.

What we measure

Five measures, one comparison.

MeasureDefinitionWhy it matters
Cases that stay solvedClosed cases that are not reopened within an agreed periodShows whether answers hold
Rework timeMinutes spent checking and correcting AI draftsShows the time really saved
Repeat contactsCustomer contacts per caseShows the customer's effort
HandoversCases handed to a person, and how fastShows whether people stay reachable
Cost per resolved caseTotal cost divided by cases that stay solvedMakes offers comparable

What to bring

Request types and monthly volumes, today’s quality checks, the team that would take part and the person who decides.