Writing archive
Older essays on agentic systems, governance, operations, incentives, and the work of making AI useful in real organisations.
My advice before you buy another AI influencer pack: try the tools on your own work, keep the instructions that actually help, and pay for demonstrated value rather than secret-prompt promises.
After bespoke development, a July Codex run and an OpenClaw detour, Grok Bot got my Chief Agentic Officer Briefing running in about two hours. The important part: it came with roughly a year's worth of learning.
A £150 million thought experiment: how wealth managers could use narrow assistants for research, records and meeting preparation, while qualified people keep responsibility and no bot trades or moves money.
How I would use Meta Muse for business preparation, customer questions and follow-through, with a small agentic family and clear limits on access.
Britain scores relatively well on the world's corruption index, but its direction of travel is worrying. Country comparisons, the UK trend and the gap between rules and enforcement.
There is no single EU boss. A straightforward guide to the Commission, Parliament, national governments, councils, courts and central bank: who decides, who votes and who can say no.
How Britain can create more useful value, rather than only argue about dividing what it already has: productivity, skills, investment, infrastructure, AI and lessons from other countries, with a test for who benefits.
A data centre keeps digital systems running. An AI factory specialises in developing and serving models. Inference, tools and permissions can help turn that capability into useful work, subject to approval and verification. Britain needs both.
UK residents can already receive income from government debt through gilts and NS&I. Who owns it, how does the income work, and could broader access help without pretending that debt or inflation risk disappears?
Britain is not uniquely failing, but its growth per person has been weak. A fair comparison with Europe and the world, separating recent results, longer-term performance and forecasts, and asking what we can learn.
How Tony would use OpenAI Dots to keep customer feedback, launches and board preparation moving, with clear sources, bounded delegation and human judgement.
How Tony would use Grok Bot for research, meeting preparation and follow-through, with a small family of connected assistants and clear human review.
AI may bring more personalised email, but agents could also turn useful messages and CCs into searchable business context. The productivity opportunity is less sorting, not less judgement.
A free business idea: start with commercial rentals around Enfield or your own local area, give AI agents trustworthy property information, and make keeping availability current the service people pay for.
Pricing Watch 4: new models offer cheaper routes, while unchanged dollar tariffs cost slightly more in pounds. A source-backed October API pricing snapshot.
What returning to the room, working on other leaders’ Challenges and hearing what happened next have added to my thinking since joining CEO CF in 2010.
Bring your experience to a company's work without handing over your whole knowledge library. Why I think AI could give consulting a new lease of life, and the practical boundaries that still matter.
ChatGPT Space, connected tools and portable plan usage could change where a company works and which software it pays for. The opportunity for small businesses, software builders and people bringing expertise without handing over their entire knowledge base.
If agents can build the implementation, the durable asset may be your company's rules, data and responsibilities. What this means for software, business costs and the people who understand how work gets done.
Gallup's country comparisons tell a more interesting story than optimists versus pessimists. Using AI, expecting it to help and trusting its answers are three different things.
If AI agents remove the hassle of buying elsewhere, Amazon has to compete on more than convenience. Product trust, delivery and whose agent we use become the real questions.
For complicated work, the useful shift is from handing an agent a task and walking away to staying close enough to see the structure, check the evidence and steer together.
AI can help us prepare, remember and follow up. The relationship still depends on listening, making useful connections and doing what we said we would do.
Before AI, who put a number on humanity's chances? Nuclear war, pandemics, climate and the algae story, with the outcomes and assumptions restored.
Privacy choice
Cookieless Matomo analytics help me understand which public pages are useful. They are on by default unless you turn them off, and your choice never affects access to the site.