- ● Live demo unavailable (Private)
- ● Architecture details
SharePoint → Voiceflow KB Sync
An automated enterprise knowledge synchronization pipeline — built to bridge Microsoft SharePoint with Voiceflow AI agents in real-time.
Problem
An enterprise Microsoft tenant (Teknicare) had deployed an internal AI assistant (Bob) to handle employee queries. The challenge: the AI's knowledge base drifted rapidly because managers had to manually re-upload documents to Voiceflow every time a SharePoint file was edited. This led to the AI hallucinating or serving stale policy information, eroding employee trust in the system.
Architecture & Solution
The synchronization engine utilizes a 7-step Make.com automation pipeline. It polls Microsoft SharePoint Online every 60 seconds, filtering for newly modified PDF and DOCX files. Approved binaries are downloaded and posted to the Voiceflow Knowledge Base API via an HTTP multipart request with explicit overwrite flags. The system then sleeps to allow asynchronous indexing before polling for a final READY state confirmation.
Implementation Decisions
- Idempotent overwrites — Forcing an overwrite flag on the Voiceflow API call completely eliminates the need for complex deduplication or version tracking logic.
- Asynchronous sleep handling — Implementing mandatory delays accommodates Voiceflow's asynchronous indexing queues, preventing false failure alerts.
- Strict filetype gating — A robust validation layer drops unsupported formats (ZIPs, spreadsheets, images) before any costly HTTP calls are made to the LLM backend.
- Audit log anchoring — Centralized scenario variables provide an immutable audit trail that can be routed to Slack or Google Sheets for operational visibility.