Insights
Notes from the work: applied AI, up close.
Four kinds of writing. Practical guides to decisions our clients face, research on what actually returns value, lab notes from inside builds, and announcements when something changes here.
Showing all 16 articles
Editor's pick · Research
Stop renting your judgement.
On one high-volume review workflow, a fine-tuned open model beat every frontier configuration we tested, at a fraction of the cost per decision.
27 Jul 2026
Practical guides
How to choose an AI consulting firm
Capability is not the differentiator any more. What to establish before you brief anyone, seven questions to ask, and the signals worth taking seriously.
4 Aug 2026
Practical guides
How to move an AI pilot into production
The work a pilot deliberately skips, the six things production adds, and the order that tends to hold.
3 Aug 2026
Practical guides
Why enterprise AI implementations fail
Seven failure modes that repeat, almost none of them about model quality, and how to spot yours in the first month.
2 Aug 2026
Practical guides
How to calculate ROI for an AI initiative
The baseline you already have, four ways value actually arrives, the cost lines that get left out, and four tests a number must pass.
1 Aug 2026
Practical guides
What an AI readiness assessment should include
Six dimensions, the questions that reveal each one, and why the output should be a shortlist rather than a score.
31 Jul 2026
Research
Your agent is not confused. It was never told.
Most pilots die on missing business context, not model quality. What it takes to treat context like code.
30 Jul 2026
Practical guides
Build versus buy for enterprise AI
Four layers, four different answers, and the third option most organisations should be using.
29 Jul 2026
Practical guides
How to evaluate an AI agent
Success rate is one number hiding four. What to measure when a system takes actions rather than producing answers.
28 Jul 2026
Practical guides
How to avoid AI vendor lock-in
Lock-in forms in what you did not keep. The five assets that keep you portable and the contract terms worth arguing about.
25 Jul 2026
Research
Most companies don't have an AI strategy. They have activity.
Adoption is common and value is rare. The four decisions that separate a strategy from a pile of pilots.
22 Jul 2026
Research
The real cost of AI isn't the budget. It's the value gap.
Five places the money leaks between the pilot and the return, and an honest afternoon that sizes your own gap.
15 Jul 2026
Research
AI ROI is an organisational problem. The model was never the bottleneck.
The same tool returns value in one company and nothing in the next. Workflow, ownership, and where freed time goes.
6 Jul 2026
Model labs
How to choose a model for production, when the best one is rarely the right one.
Frontier models are close on quality and far apart on cost. What we test instead of the leaderboard.
24 Jun 2026
Announcements
CoSeeker is now open source.
A contemplative social platform on the AT Protocol, with its code now open. What we are releasing, and why.
10 Jun 2026
Announcements
We acquired a CEO coaching practice.
Radical Results joins K4M2 AI. The models were never the part that stalled, the leadership team around them was.
19 May 2026