Blog

Field notes from AI activation

Original essays on what actually happens when teams adopt AI (the methodology behind the AI Activation Sprint, written from the workshop floor, not the keynote stage).

An experienced guide points along a gold bridge crossing five successive technology waves toward a sixth illuminated opening

AI Adoption • 19 May 2026 • EN

Five technology waves, one repeating pattern: what 26 years of transitions predict about your AI program

Twenty-six years across five technology transitions, and the same failure every time. Why organisations keep treating each wave as a technology problem when it has always been a clarity problem.

Five professionals open forms that reveal the gap between AI promises and the reality of the work

AI Adoption • 12 Sept 2025 • EN

5 counter-intuitive lessons from AI training and coaching

Five lessons from supporting dozens of teams through AI activation: the opposite of what the market keeps promising.

A woman turns an AI form over to reveal its blind spots and missing pieces

Team Dynamics • 21 Aug 2025 • EN

What if we held AI upside down?

A reversal of perspective: what if AI is not a tool to master, but a mirror that reveals our cognitive blind spots?

A team passes context through a shared work relay

Team Dynamics • 30 Jul 2025 • EN

Does AI force us to rethink how teams work?

When every member of a team has a copilot, roles, rituals, and work distribution begin to deform. How do we recompose them?

Stewards build a finite compliance wall while a leader opens a clear path for colleagues standing in organisational fog

Commentary • 21 Jul 2026 • EN

The EU AI Act won't kill your AI program. Fog will.

Compliance is a bounded problem you can schedule and retire. Fog is not. Why waiting for legal clarity is often paralysis wearing a compliance badge, and the two questions that tell them apart.

A strategist aligns goal, context, permissions and verification into one guidance rail shared by a team and an agentic work instrument

Technical • 14 Jul 2026 • EN

Your AI agents don't have a goal problem. Your organization does.

Hand your corporate objectives to an AI agent and it fails instantly, for reasons an engineer can name precisely. The same reasons your people have been quietly absorbing for years.

Five stakeholders actively assemble their partial organisational fragments into one illuminated shared map

AI Adoption • 7 Jul 2026 • EN

The Complexity Map: seeing where your organization is actually stuck

IT says legacy, business says IT, leadership says culture. Everyone is partly right, which is the problem. Why the first deliverable of any engagement is a map, not a strategy.

A prompt chamber sits inside a context library, bounded workbench and evidence loop leading to a verified completion gate

Technical • 30 Jun 2026 • EN

From prompts to loops: the four-layer shift your AI strategy hasn't caught up with

Prompt engineering, context engineering, harness engineering, loop engineering. Each layer vendors sell as a technical specialty is an organisational discipline your company needed before AI existed.

Below a stalled mechanism, a facilitator helps an experienced sceptic turn a hidden brake into a calibration instrument

Team Dynamics • 23 Jun 2026 • EN

Some of your employees are quietly sabotaging your AI rollout. They have their reasons.

When employees quietly undermine an AI rollout, the boardroom asks who. The useful question is why. Resistance is feedback that could not find a safer channel, and it is readable.

A workflow owner opens a threshold so one pilot leaves a suspended test enclosure and enters real operational work

AI Adoption • 16 Jun 2026 • EN

Pilot purgatory: why most AI pilots never reach production — and the one question that gets you out

Most AI pilots are designed to avoid the organisation rather than change it, which is exactly why they never cross into production. The one kickoff question that predicts whether yours will.

A team opens an obsolete operating machine and builds one end-to-end flow behind two generations of bolt-on technology

AI Adoption • 9 Jun 2026 • EN

The mistake organizations made with mobile in 2012 is the one they're making with AI in 2026

Replace mobile app with AI pilot and 2012 becomes 2026. A technology that changes how work flows cannot be adopted as an accessory to work that has not changed.

Leaders leave a crumbling linear plan and navigate using a stable direction, one illuminated step and a living calibration wheel

AI Adoption • 2 Jun 2026 • EN

Your AI roadmap is a lie. Here's what you actually need.

A three-year AI Gantt chart is not a plan, it is anxiety management. What replaces it: a direction people can see, the next verifiable step, and a cadence of honest re-decision.

A facilitator lowers three illuminated probes into ignored decisions, unchanged daily work and an unspoken organisational silence

AI Adoption • 26 May 2026 • EN

Three questions I ask in the first week of any transformation engagement

What have you decided that nobody is doing? Whose Monday morning changes? Where do people go quiet? Three questions that locate where an organisation is actually stuck, faster than three months of audit.

A man moves essential context from an overloaded vessel into a clean conversation space

AI Literacy • 2 Oct 2025 • EN

When should you start a new AI chat, and when should you continue?

The context window is a finite resource. Knowing when to reset and when to carry on is one of the highest-leverage skills in AI literacy.

Two people open a fluent answer to inspect the unsupported elements inside it

AI Literacy • 18 Sept 2025 • EN

How to spot a hallucination before it spots you

Five practical tells that an AI answer is fabricated, written for non-technical readers who want to trust their tools without being burned by them.

A strategist adjusts a frame that turns an ambiguous request into precise results

AI Literacy • 4 Sept 2025 • EN

Why the way you write your prompt changes everything

A non-technical deep-dive into attention, context, and why the same model gives radically different answers to almost-identical questions.

Three languages travel unequal tokenisation paths to reach the same meaning

AI Literacy • 12 Aug 2025 • EN

Not all languages are equal for AI

Tokenization is not neutral. French costs more tokens than English for the same idea, and that has direct consequences for cost, latency, and quality.

A practitioner cleans, chunks, and organises documents before retrieval

Technical • 25 Jun 2025 • EN

Document preparation: the key performance factor in RAG systems

In a RAG project, 80% of final quality is decided before the model: in how sources are cleaned, segmented, and indexed.

An analyst compares three models using equal inputs and different outputs

Commentary • 30 May 2025 • EN

Grok 3, O3, and Claude 4: an in-depth face-off

A methodical comparison of three frontier models on real intellectual work tasks: not a benchmark, but field use.

A strategist extracts three useful signals from a vast report while keeping a blind spot visible

Commentary • 15 May 2025 • EN

Mary Meeker's 2025 AI Report: what to take away

A selective reading of the 340-page report: three charts that really matter, and one blind spot worth naming.

One developer directs an instrument that extends their reach across product, design, engineering, and delivery

Commentary • 22 Apr 2025 • EN

The total developer, or the age of versatility

An adaptation and commentary on Justin Searls' essay about the announced disappearance of rigid specialization in software work.