I've been evaluating the new HubSpot integration for MeetGeek over the last two weeks, primarily to see if it can reliably push meeting transcripts and summaries into our CRM without manual intervention. The promise of automated, structured data flow is compelling, but the devil is in the implementation details.
From a technical standpoint, the integration uses OAuth 2.0 for authentication and exposes a webhook endpoint for meeting events. The core workflow is straightforward:
1. Meeting ends in MeetGeek.
2. A JSON payload containing `transcript`, `summary`, `attendees`, and `action_items` is sent to a configured HubSpot contact or deal record.
3. Properties are mapped in HubSpot.
However, I encountered several friction points:
* **Property Mapping Limitations:** The mapping interface only allows top-level field assignment. Nested data from the action items array, for example, cannot be cleanly split into separate HubSpot properties without custom middleware.
* **Payload Consistency:** In three test meetings, the `duration` field was missing from the webhook payload twice, despite being present in the MeetGeek UI. This suggests an inconsistency in their internal API.
* **Error Handling:** Failed syncs (due to HubSpot API rate limits in our case) generate a generic error in MeetGeek's log but no automatic retry mechanism is documented.
The integration "works" for basic use—summary text does appear in a HubSpot note. But for engineering teams expecting a robust, production-ready data pipeline, it feels undercooked. The lack of idempotency keys and retry logic means we cannot guarantee data delivery.
Has anyone else stress-tested this integration with a high volume of meetings? I'm particularly interested in:
* Your experience with the webhook delivery latency.
* Whether you've built a middleware proxy to handle data transformation or retries.
* Any workarounds for the property mapping constraints.
benchmark or bust
benchmark or bust