All field notes

Opinion

·

10 min read

From chatbot to coworker: the state of AI operations in 2026

Agents have moved from answering questions to doing work. What changed, and what still doesn’t work.

Maya Okafor

Co-founder & CEO

Two years ago most companies asked us for a chatbot. This year, nobody did. They asked for agents that take actions: update the CRM, rebook the shipment, reconcile the account.

What changed

  • Models got reliably good at using tools

  • Costs fell enough to run agents on every transaction

  • Teams learned to design oversight instead of avoiding risk

What still doesn’t work

Agents struggle with work that has no clear definition of done, with processes that change weekly, and with data nobody trusts. No model fixes a messy process — it just automates the mess faster.

Fix the process, then automate it. In that order.

What’s next

We expect the next year to be about coordination: several specialised agents handing work to each other, with humans managing the system rather than the tasks.

Written by

Maya Okafor

Previously led operations at a Series C logistics company. Has automated her own inbox three times.

Keep reading

Put your operations on reflex.

Tell us where your team loses the most time. We’ll show you what an agent could take off their plate — and what it would save — in one call.

Create a free website with Framer, the website builder loved by startups, designers and agencies.