Field notes from
the workshop.
Things we've learned building agents in production. Honest opinions, useful patterns, occasional rants.
- October 10, 2026
The support agent that knows when to escalate
When an AI support agent should hand a ticket to a person: the hard rules, the soft signals, and why "I want a human" always wins.
- c4-support
- escalation
- customer-support
- October 7, 2026
What an AI BDR agent actually does, end to end
The seven steps an AI BDR agent runs on every lead, from trigger to booked meeting, and where it breaks: stale data, send volume, and the law in the inbox.
- c2-pipeline
- ai-bdr
- inbound-leads
- October 4, 2026
Rolling out an AI agent: shadow mode, supervised, autonomous
The three stages we roll an AI agent through, who on your side should decide each move, and what you watch while it runs.
- c8-trust
- rollout
- shadow-mode
- October 1, 2026
What it costs to build and run an AI agent
The three cost lines nobody quotes: the build, the run, and the cost of the agent being wrong.
- c7-engagement
- cost
- scoping
- September 28, 2026
How to build an AI agent: the decisions only you can make
Five decisions in an AI agent build belong to the business: the job, what you hand over, the owner, your team's time, and the scope you sign.
- c7-engagement
- scoping
- sign-off
- September 1, 2026
Should this workflow be an AI agent? A five-question diagnostic
The five questions we ask before agreeing to build an AI agent - including the two that produce a no. Run your own recurring workflow through them first.
- c1-diagnosis
- ai-agents
- scoping
- automation