Articles & series
Fewer Moving Parts
Making work simpler without moving the problem.
Practical articles about systems, processes, people and technology, with the costs, constraints and operational detail left in.
Subscribe on LinkedIn ↗Series 01
Should We Automate This?
Decisions before demonstrations.
Four decisions for defining the work, testing suitability, checking the project case and designing a pilot allowed to say no.
Part 1 of 4
01How to Define an AI Project Before Choosing the Technology
AI projects should begin with a measurable business change, not a product or model. This guide provides a five-question test, a process map and a one-page opportunity brief.
Read article →Part 2 of 4
02How to Decide Whether a Business Process Is Suitable for AI
Repetitive work is not automatically suitable for AI. Use this six-part assessment to examine the work, its exceptions, its evidence and the cost of being wrong.
Read article →Part 3 of 4
03How to Assess the Business Value, Risk and Readiness of an AI Project
A technically suitable AI use case can still make a poor project. Assess value, risk and readiness together before approving a pilot.
Read article →Part 4 of 4
04How to Design an AI Pilot That Produces an Honest Decision
An AI pilot should test a business decision, not extend a demonstration. This framework covers baselines, case selection, staged exposure, measures and stop criteria.
Read article →Series 02
Local AI, Real Work
Control is useful. Responsibility moves with it.
Four practical tests for deciding when local AI is capable enough, what the surrounding system must provide, how configuration changes the result and when local deployment earns its place.
Part 1 of 4
01When Is Local AI Good Enough for Business Work?
Open-weight models can now handle credible business work on infrastructure an organisation controls. The decision still depends on the whole workflow, not the model alone.
Read article →
Part 2 of 4
02The Seven Parts of a Working Local AI System
Model selection is only one part of local AI. A working service also needs controlled information, tools, permissions, review, records and recovery.
Read article →
Part 3 of 4
03How to Test Reasoning Settings for Business AI Work
In one quotation test, no reasoning failed, a low setting produced the strongest result and more reasoning made it worse. The configuration needs its own evaluation.
Read article →
Part 4 of 4
04Local vs Hosted AI: A Seven-Gate Decision Framework
Local AI offers control but transfers operating responsibility. Seven decision gates help determine whether local, hosted or hybrid AI fits the work.
Read article →