Featured · AI in the Organization
Start with the queue, not the chatbot
A practical way to choose an AI use case by following delayed work, measuring the constraint, and testing the smallest useful intervention.
Read the articleInsights
Practical decisions for founders and operators: where to start, what to build, and how to make the change hold after launch.

Featured · AI in the Organization
A practical way to choose an AI use case by following delayed work, measuring the constraint, and testing the smallest useful intervention.
Read the articleBrowse by question
Choosing the work, setting boundaries, and helping a team change how it operates.
Finding the constraint before deciding whether software deserves to exist.
Architecture, evaluation, and the unglamorous work that makes AI reliable.
Latest notes
10 guides on software, AI, and changing the work around them.
AI in the Organization
What Iceland's teacher pilot can teach SME leaders about AI adoption: start with practitioners, one real workflow, explicit boundaries, and a learning loop.
From the Build
A practical agent-readiness guide covering process boundaries, permissions, idempotency, final-state checks, and evaluations before adding autonomy.
From the Build
A practical evaluation method for AI workflows that measures useful outcomes, dangerous failures, human review, adoption, speed, and cost.
AI in the Organization
Trace AI value from real use to changed workflow behavior, operational results, and financial outcomes with a practical measurement card.
Better Business Systems
A practical way to tell whether the spreadsheet is the constraint—or whether the real problem is ownership, handoffs, and workflow design.
AI in the Organization
A one-page AI policy starter for SMEs covering approved tools, data boundaries, review, prohibited decisions, incidents, and ownership.
Better Business Systems
A practical discovery guide for tracing one real work item, finding the constraint, and defining a first release without writing a technical brief.
From the Build
Choose between rules, a model call, a bounded AI workflow, and an agent by matching autonomy to variation, risk, and evidence.
Better Business Systems
Trace one real piece of work to decide whether to configure, connect, build custom software, or leave the process alone.