The €40,000 question: how to calculate whether an AI automation is actually worth building.
A line-by-line ROI model — labour cost, error cost, opportunity cost, decay — used internally on every project we accept. Includes the spreadsheet logic and the failure modes most calculators miss.
Why your last AI tool got abandoned in 30 days — and the four onboarding moves that fix it.
Adoption is not a training problem. It is an interface, incentive, and trust problem. Here's the four-move playbook we run on every shipped agent — and the cost of skipping any one of them.
Lead response time benchmarks for SMBs in 2026: what's normal, what's possible, what it costs you.
Anonymised data from 41 engagements. The median is worse than you think; the top decile is closer than you think. Includes the back-of-envelope formula for what every minute of delay costs your pipeline.
A line-by-line breakdown of an internal knowledge assistant we built for a 40-person professional services firm.
Architecture, sources, permissions, citation rules, evaluation harness, and the seven failure modes we caught before launch. The most-detailed piece on this site, written for the operator who has to make this decision.
The five workflows we automate first in almost every SMB — and the three we deliberately leave alone.
Pattern recognition from 184 engagements. Some workflows look automatable but aren't. Some look small but compound. Here are both — with the reasoning that drives the call.
The "human-in-the-loop" pattern, properly. A reference architecture for AI systems your compliance team can sign off on.
When to escalate, who to escalate to, what they see, what they don't, and how to keep the model from drifting around the gate. The pattern that keeps regulated industries shipping AI in production.
The SMB Automation Brief.
Bi-weekly. One real case study with numbers. One mistake we're seeing. One tool worth knowing about. 8,400 operators reading. No funnels.