Agents that run the company

AI agents for data entry, support, alerts and analysis in a small company, each with a human owner and a human decision before anything customer-facing changes.

Problem

Event data entry, testing and alerting took manual hours every week.

Constraints

Agents touch live customer data, so a wrong venue field is a public mistake. Every run has a cost limit, and a small team cannot babysit a bot.

Approach

Each agent gets a named human owner. Agent Board reads venue pages and proposes each field with its source, and a person decides every field before it goes live. A WhatsApp customer-support agent answers from trained replies with an AI fallback, steps in only when nobody replies in time, and hands refunds and complaints to a person. Error alerts turn every failed request into a plain-language message with a help code.

Outcome

• 6 agents in daily use, covering alerts, IT automation, ops briefings, data import and social analytics.
• Agent Board live in production since October 2026.
• Customer-support agent built and deployed, in testing, not yet answering real guests.

Still open

Turning on the support agent in a small, supervised pilot, then widening it once the answers hold up.

Build a product · Advisory · CTO as a service