WhatsApp AI Workflows That Actually Ship
A practical playbook for shipping WhatsApp-based AI assistants that handle intake, qualification, and follow-up — without the chatbot-demo feel.
WhatsApp is where a huge portion of the world's business conversations already happen. A well-designed AI workflow meets customers where they are and quietly handles the boring 80% — leaving humans to focus on the 20% that actually needs judgement.
The workflow shape we keep coming back to
- Intake — capture intent in natural language, not forms
- Qualify — ask only what's needed to route the next step
- Act — book, quote, answer, or escalate
- Follow-up — nudge, remind, close the loop
Things that go wrong
- Too many clarifying questions — users drop off
- No human handoff — hard cases stall
- Prompt-only logic — hard to test, hard to evolve
The fix is boring engineering: structured state, tool calls, evals, and a clean operator UI for the exceptions.
Related posts
CI/CD pipelines are deterministic. AI agents can reason. Do we still need pipelines when agents can choose what to run and triage failures — or is determinism the one thing agents can't replace?
E2E testing is shifting to natural-language AI agents while integration tests still need deterministic code. Where does each layer of testing belong — and do we actually agree on what 'right' looks like?