Managers Already "Vibe Code". We Call It Management.
Briefing, delegation, judgement and checking are familiar management skills. They offer a useful starting point for working with AI, not permission to stop learning or reviewing.
Agentic AI & Community Building
Tony Wood.
I help leaders put agentic AI into practice and build communities where people learn from each other. Advisory, talks and workshops grounded in boards, operations and CEO peer learning.
Work with Tony
Useful conversations. Practical next steps.
For boards, leadership teams and communities working through what AI changes, what stays human, and what to do next.
Latest writing
Shorter essays and operating notes
Could Britain Make Its Own Fuels?
Britain can manufacture hydrogen, fuel-cell systems and synthetic fuels. The opportunity is to match each to a useful job, with credible electricity, feedstocks, customers and costs.
Could AI Give Us More Time for Our Communities?
AI could help people reclaim time and charities organise it. A community revival would need income security, flexible roles and funded human support, not just more technology.
Popular Articles
The articles attracting the most interest at the moment, and my take on why.
Why Do We Work?
We work to pay for life. But a salary, a sense of purpose and a good life are different things. Could AI help us separate them?
Tony's Take
I think this gets people because it's a question we don't often stop to ask. We've got bills to pay, of course. But if the money were sorted, what would you actually choose to do with your time? All this talk about AI taking jobs makes that worth thinking about. There's more to a good life than having a job.
I Thought It Was Cultural Etiquette, Not AI Etiquette
A new inbox label for a familiar irritation: AI wrote the reply, but did the sender read it? What mobile-phone etiquette can teach us about judgement, shared attention and the manners of AI.
Tony's Take
My guess is people recognise this. Somebody puts your message into AI and sends the answer straight back without reading it. Now you've got to make sense of something they haven't bothered to understand themselves. Come on. Read it first. I think we're still working out the manners of using these tools with other people.
Was Merlin Real?
There is no secure evidence for a historical wizard called Merlin. There is much better evidence showing how medieval writers assembled the legend.
Tony's Take
This started with me wondering what deep research could do with a question like "Was Merlin real?" I suspect the curiosity is part of the appeal. You know the name, you know some of the stories, but where did they actually come from? Following that question was interesting in its own right. We're allowed to use these tools just to learn something.
Research
Playbooks, papers, and board-level frameworks
Agentic Language: The Common Language Layer For Agentic Work
Why read this
Start here when agents and people need shared terms for requests, commitments, evidence and handoffs, not just more conversation.
A research paper proposing agentic language as a common operating layer for humans and agents: shared business terms, Agent Moves, Meaning Blocks, receipts, judgement, human-care signals, and pattern promotion.
Consensus Is Not Compliance
Why read this
Use this to distinguish a decision people have tested and can support from agreement produced by pressure or silence.
A research paper proposing consensus as a disciplined process of conviction, evidence, listening, challenge, adaptation, and commitment, then translating that human governance model into agentic systems.
Start here
Find your next read.
Start with agents
One useful workflow, clear boundaries and a first step you can test.
Lead and govern AI
Keep judgement, ownership and evidence visible as AI takes on work.
Make AI useful at work
Move from an impressive demo to work someone can rely on.
Build communities
Create space for people to challenge, contribute and learn together.
Understand costs
Distinguish a model price from the cost of useful, completed work. Monthly snapshots are dated, not live tariffs.