Case studies
A contact-centre analytics company
83 agents and 42 workflows that score conversations and answer data questions, called from their own product.
Who they are
A company whose platform analyses contact-centre conversations.
Shared without the customer’s name.
The problem
- Their platform scores contact-centre conversations, raises alerts and reports on agent performance.
- That takes many specialised AI tasks, at volume, at low cost.
What runs on Studio
83 agents and 42 workflows, called from their own product through Studio’s API.
- Scoring agents, such as guideline scorers, an emotional-intelligence scorecard, a satisfaction forecaster and risk radars.
- A data copilot that answers questions with SQL tools.
A tested model switch
- 01
They compared models on the copilot’s test suite in Studio.
- 02
A smaller model scored 26 of 27 cases across three runs.
- 03
The cost of a copilot turn fell from about $0.05 to $0.001–0.003.
- 04
The previous version stays one click away.
The numbers
- agents
- 83
- workflows
- 42
- sessions since September 2026
- 38,000+
Figures from the customer’s production Studio workspace, November 2026.
Why it matters
An AI product built on Studio, not a chatbot. And a tested model switch cut a key cost by 20–50×.
Start building with Krazimo
Build with confidence.
Deploy with evidence.
Define your guardrails, run your evaluations, and review the results. Bring your agents and workflows to production with the checks in place.