Retell AI webhook data not mapping through Make AI Toolkit to downstream modules

Hi Joseph,

This looks like a field-shape issue between the Retell webhook, Make AI Toolkit, and the downstream modules rather than a rebuild problem.

I would trace one redacted execution through three checkpoints:

  1. The raw Retell bundle
  2. The AI Toolkit output and whether it is an object or a JSON string
  3. The exact field paths expected by Calendar, Outlook, Twilio, and Sheets

Then I would normalize a small agreed set of fields such as name, phone, email, appointment time, and summary into fixed Make variables, prove one downstream mapping, and return a short before and after note. I would not need credentials or real client data to confirm the first failure.

If you still need help, I can take this as a contained async troubleshooting pass rather than a rebuild.

@Joseph_Linco
The useful detail here is that Retell is sending the right values into the webhook. That narrows the failure to the mapping or type changes between the webhook, AI Toolkit, and the first downstream module.

I would repair this in place. First, capture the raw webhook bundle and the AI Toolkit output side by side. Then trace one client record through every mapping boundary, correct the schema mismatch, and test the appointment, email, SMS, and Google Sheets paths with the same record.

Is the first bad value already missing in the AI Toolkit output, or does it disappear when the next module maps the bundle?

Happy to help. Feel free to reach out.
Paul

Hi Joseph β€” this looks like a field-contract/mapping issue rather than something that needs a rebuild.

I’d compare one redacted Retell webhook bundle against the AI Toolkit output, identify where the client fields change shape or disappear, normalize the required fields into a stable structure, then prove the fix through one downstream module before retesting the complete flow.

My strongest work is Make/API workflow architecture, deterministic mappings, error tracing, and defect β†’ root cause β†’ patch β†’ retest.

If this is still open, I’d be happy to look at a redacted failed execution and the AI Toolkit output. No credentials or real client data needed.

You can also reach me directly at [email protected] if that’s easier.

Best,
Valerii