Question 1

Will you use n8n or a custom Python pipeline?

Accepted Answer

It depends on the volume and complexity. n8n is the right call for most workflows under ~10,000 events per day — faster to build, easier to hand off. Beyond that, or where you need bespoke logic the available nodes can't express, I drop into custom Python with a state machine in Postgres. The decision goes in the brief in week one.

Question 2

Can the workflow handle webhooks from anywhere?

Accepted Answer

Yes. Most builds expose a webhook URL on n8n that any third-party service can POST to (Zendesk, HubSpot, Slack, Stripe, Calendly, custom forms). Authenticated with a header secret. Replies asynchronously so the calling system never blocks.

Question 3

How long does an autonomous-support routing build take?

Accepted Answer

Nine days end-to-end for the median case: intake → classification → reply drafting → routing to the right inbox. That includes a clickable prototype in week one and the production cutover in week two. Real numbers from a Series-B fintech build.

Question 4

What happens to my data — does anything leave the box?

Accepted Answer

OpenAI processing leaves the box (that's the trade for using their models). Everything else stays inside the workflow's Postgres. Data retention is configurable per node. If you need fully on-prem, I swap OpenAI for Llama or Mistral running locally.

AI & Business Automation

Key Capabilities

Need help with AI?