AI Agents
AI churn-risk detection from Microsoft Teams check-in transcripts
A daily Azure pipeline that scores every coach–VA check-in meeting against a 22-signal churn-risk checklist and lets leadership ask questions like “Which VAs are high risk this week?” in plain language.
- Client
- Virtual assistant staffing agency
- Industry
- Professional services
- Status
- In production
The problem
Early warning signs such as burnout, disengagement or client dissatisfaction were buried in hours of Teams transcripts. There was no consistent way to compare risk across assistants and clients.
What we built
- Timer-triggered Azure Function pulls the last 7 days of meetings and transcripts through Microsoft Graph, with Azure Speech as a fallback
- GPT-4.1 scores each transcript 0–100 against a 22-signal checklist (Low, Medium, High, Critical) and writes coaching recommendations
- Azure AI Search hybrid index plus an Azure AI Foundry agent answer natural-language questions
- Power BI dashboards and a SharePoint feedback loop that improves scoring through reviewed examples
Challenges we solved
- Meetings with recordings but no transcript
- Parsing coach, VA and client names from inconsistent filenames (40+ name mappings)
- Azure OpenAI rate limits during batch analysis
Results
- Running daily in production since v2.0, upgraded in v2.1
- Every meeting scored and searchable; risk visible in Power BI and through the agent
Technology stack
- Azure Functions (Python)
- Microsoft Graph
- Azure OpenAI GPT-4.1
- Azure AI Search
- Azure AI Foundry
- Azure Speech
- SharePoint
- Power BI
Need something similar?
Free 30-minute scoping call. We reply within 24 hours.
Talk to us about AI Agents