Your Inbox Is the New Interface: Building AI Agents on Chat applications

About this session

Every AI startup deck in 2026 promises an agent. Almost none of them have a distribution channel. Meanwhile, businesses are already having millions of conversations a day on WhatsApp — and drowning in them.

This talk maps the AI startup landscape through a lens most founders overlook: the business messaging layer. Using WhatsApp Business API and the Meta Business Agent as a working example, I'll show how customer conversations became the highest-intent, lowest-friction surface for AI products — no app install, no onboarding, no behaviour change required.

Drawing on hands-on experience supporting enterprise messaging deployments at Meta, I'll cover: what actually breaks when you put an LLM in front of a real customer inbox; the difference between a chatbot and an agent that can transact; the gaps in today's ecosystem where startups can credibly build (routing, QA, compliance, multilingual support, human handoff); and where incumbents will close the door on you.

Attendees leave with a concrete map of buildable opportunities in conversational AI — and a realistic view of the failure modes nobody puts in the pitch deck.

Speaker

Key takeaways

  • Why messaging is the highest-leverage distribution channel for AI products in 2026, with adoption numbers to back it
  • A map of unbuilt/underbuilt opportunities in the conversational AI stack, and which ones are platform-risk traps
  • A live look at what breaks when an AI agent meets a real business inbox — latency, hallucination, escalation, compliance

Related sessions