AI Return Model by OnClarity What happens when support fixes things the first time.
The OnClarity AI Return Model follows seven personas through 48 simulated months of a contact centre, hour by hour. Every number is modelled, sourced and stress-tested.
What changes in a reference contact centre after the switch?
Seven personas, one queue.
Everyone in a contact centre shares the same queue. Change what goes into it and all seven personas change.
Why is contained not the same as resolved?
A generic AI that is paid per contained conversation counts give-ups as wins. They come back as new contacts. OnClarity counts an issue as resolved only when it doesn’t come back.
| Generic AI | OnClarity | |
|---|---|---|
| Claimed contained | 80% | 79% |
| Verified resolved | 66% | 78% |
| Issues never resolved | 16.0%, worse than today’s 15.1% | 5.2% |
| Agent attrition | 53% | 29% |
Which decision changes the outcome for agents?
The AI frees capacity. How you staff the team after go-live decides how much of it reaches agents and waiting customers. Between 65% and 80% planned occupancy, every persona wins.
How does the AI Return Model work?
Eight connected layers, run month by month for four years, with an hourly queue inside every month.
Evidence behind the inputs
30 inputs: 11 from peer-reviewed or analyst research, 10 client or OnClarity inputs a pilot replaces, 9 ranged assumptions.
Stress test
100 simulations with 22 inputs varied at once. With balanced staffing, all seven personas are better off in every one.
3-year value in the reference organisation: $5.4M to $13.8M (10th to 90th percentile). Reference run $12.8M, marked in white.
Sources
- Brynjolfsson, Li and Raymond (QJE 2025)
- Brown et al. (JASA 2005)
- Gartner (2019, 2023, 2024, 2025, 2026)
- SQM Group
- McKinsey
- Forrester
- Bakker, Demerouti and Schaufeli (2003)
- Keiningham et al. (2007)
Run it on your contact centre.
Four numbers. About twenty seconds. A report and a PDF your team can check line by line.
Questions a CFO will ask.
It simulates the operation month by month: queues, repeats, attrition, hiring, churn. Cost and value come out of that simulation instead of being assumed.
All figures on this page come from the OnClarity AI Return Model run on a reference organisation of 1 million customers and 200,000 issues a month, at steady state. They are model outputs, not forecasts or guarantees. Third-party statistics are cited with their sources.
