CUSTOMER ACCOUNT / ATTRIBUTED EVIDENCE
What this case establishes
This is a commercial publisher’s customer account. Roseram checked the source and wrote the analysis; it did not interview the customer, run the product, audit the figures or establish that AI caused the reported business result. The pilot below is a proposal, not a completed experiment.
Primary material reviewed 2026-10-12. Analysis and proposed evaluation by Roseram Editorial.- 01Eligible customer question
- 02Grounded answer or escalation
- 03Verified task outcome
Intercom: the documented workflow
Anthropic’s Intercom account describes Fin answering support questions using Claude. It reports an average initial resolution figure and higher outcomes for some deployments. “Up to” and “average” describe different populations; neither should be used as a guarantee for a new business or as proof that every unresolved customer received an adequate handoff.
The mechanism worth examining
Resolution is meaningful only when its denominator and ending condition are defined. Was the case eligible for automation, answered once, accepted by the customer or closed without another contact? A useful support system can resolve routine questions and transfer exceptions with context. The business should optimize the customer’s completed task, rather than teach the bot to close conversations quickly.
The implementation decision
Choose a support category with reliable approved documentation. Define excluded cases and escalation ownership first. Keep a human route visible, especially for account access, complaints or decisions the bot cannot make. Compare automation on similar ticket types; adding easy questions to the denominator can improve a headline while leaving the difficult customer journey unchanged.
Where the workflow can fail
A silent customer can look resolved even after giving up. A repeat contact may arrive through another channel and be counted as a new unrelated event. Link follow-up behavior where permission allows and sample closed cases for actual task completion. Do not allow a confident answer to bypass identity checks or promise a refund outside approved policy.
A practical evaluation, designed for this use case
Use a labelled historical sample and a controlled live pilot with an accessible human fallback. Track eligible cases, supported answers, repeat contacts, escalation delay and customer completion separately. Include unclear documentation and missing account context. A credible report explains what the bot handled and what remained with people instead of advertising one blended percentage.
Primary sources
Figures and product details above are dated and attributed; they are not current guarantees.
- Anthropic: Intercom primary account ↗The attributed workflow and reported result. Proposed pilots and decisions below are original Roseram analysis.
The decision matrix
Original evaluation criteria for this workflow. These are questions to test, not results measured by Roseram.
| Decision | Evidence to collect | Stop or revise when |
|---|---|---|
| Resolution definition | Denominator and ending condition stated | Silence is treated as satisfaction |
| Escalation | Context reaches a human owner | Customer must restart the case |
| Coverage | Eligible ticket categories defined | Easy cases inflate the headline |
YOUR EVALUATION / NOT A REPORTED RESULT
A small pilot you can inspect
Use these steps as a planning checklist. Checking a box records your progress in this view; it does not run or validate an experiment.
Questions this case answers
Is “up to” a forecast for my business?
No. It identifies an upper-end reported result in a particular deployment.
What should a support pilot include?
Supported answers, exclusions, escalation and repeat-contact outcomes, with a clear denominator.
Tools & capabilities referenced
Operators studying this story often evaluate adjacent AI products. Explore Roseram guides where available:
PRACTICAL NEXT STEPS
Take the evidence into your own project.
Calculate the whole project cost
Compare subscriptions, usage, launch work and renewals.
Test an existing-project workflow
Check import, source loading, preview and publication.
Verify the output you buy
Require evidence tied to a real customer task.
Wiki brief
Intercom is examined here for resolution denominators and human escalation. The article answers “How should a business measure AI customer-support resolution?” using linked primary material and original evaluation design. It is not a customer testimonial for Roseram or an endorsement by Intercom.
Operator takeaways
- Define resolution precisely.
- Keep a usable human route.
- Inspect repeat contacts.
Related case studies
Plan → build → review with routed models and transparent usage.