The AI pilot worked. Should it go into production?
The demonstration ran on selected cases with the project team close by. A production budget has to cover the conditions the trial did not exercise.
Read the articleFeatured
Current analysis on funding, production, and continuation.
Each article starts with the decision, the available evidence, and the commitment being requested.
Review the AI business case before funding the initiative
Model capability settles one claim in the AI business case. The budget still rests on customer behaviour, staff workload and costs that sit outside the prototype.
Read the articleA delayed software program needs a continuation decision, not another forecast
The revised date explains the current schedule. The funding decision depends on the programme that exists now and the value still available.
Read the articleNotes
Questions that carry from the demonstration into the operating plan.
What an AI prototype proves before production
A prototype can establish technical feasibility under selected conditions. Production readiness requires evidence about ordinary use, operating cost, support, reliability, and control.
Read the articleDefine the operating boundary of an AI agent
An agent with tool access needs a defined role, a controlled workspace, visible limits, and accountable oversight.
Read the articleConstructive skepticism improves technology investment decisions
Skepticism works as a circuit breaker when it turns concern into a test, a funding condition, or a limit on exposure.
Read the articleArchive / 2023–2025
Earlier writing, preserved in its original context.
The original URLs and publication dates are retained. Pieces about models, platforms, and policy reflect their publication context.
Archive / AI policy
The AI arms race, viewed from February 2025
I wrote this when the public AI debate appeared to be moving from guardrails towards state competition and military investment.
Open archive entryArchive / AI systems
OpenAI o3 and DeepSeek R1: two strategies in February 2025
The piece compared two strategies through the access, control and tool-use choices visible at the time.
Open archive entryArchive / Media
The Blurred Lines of Reality
A familiar news format can create confidence in a clip even when the event behind it never happened.
Open archive entryArchive / Future scenario
A day in 2027: where does reality end and AI begin?
One fictional day follows a person whose environment predicts each preference before it feels fully formed.
Open archive entryArchive / Immersive learning
What repeat play at PLAYCON showed me about immersive learning
Children returned to the marine-biology lesson to improve their scores, giving the team a direct view of play and learning in the same session.
Open archive entryArchive / Decision quality
The Hidden Flaws of Reasonable Decisions
Coherent reasoning can preserve a faulty starting point all the way to a confident conclusion.
Open archive entryArchive / Speaking
National Skills Council panel: Emerging Technologies as a Catalyst for Innovation
The surviving entry records the panel topic and my role as a speaker.
Open archive entryArchive / Speaking
SOHO Talks: Breaking Barriers
Only the event title and publication date survive in the public archive.
Open archive entryArchive / Emerging technology
The Era of Epochal Tech Revolutions
I revisit an early dismissal of AI and ask when skepticism stops testing a claim and starts ignoring new evidence.
Open archive entryArchive / Digital economy
X’s creator revenue model, viewed from 2023
The article examined a platform paying creators from advertising income generated around their participation.
Open archive entry