How I built LeadDuo’s agentic workflow using Inngest + OpenAI (architecture & micro-agents)

Building LeadDuo’s real-time engine was the hardest and most fun part. I used Inngest to orchestrate all the background workflows that happen the moment a form is submitted — scoring the lead with OpenAI, storing metadata, updating the pipeline, and pushing real-time changes via Upstash Redis. Everything runs on Vercel + Next.js, which gives me fast edge functions and a clean dev experience as a solo founder.

I also built separate AI “micro-agents” that handle specific jobs: one agent scores leads, another triggers notifications (Slack, Airtable, Notion, webhooks, SMS, email), and a third replies directly to the form submitter. Each agent runs as its own Inngest function, which keeps things reliable even when multiple integrations fire at once.

If anyone’s curious about building agentic workflows with Inngest, Next.js, OpenAI, and Upstash Redis, I’m happy to share code snippets or go deeper. This setup lets LeadDuo feel instant for users — even though a lot is happening under the hood.