Writing
-
Note
When Services Never Stop Changing
Government has strong machinery for funding and governing programmes of change. But what happens when the thing we really need is a service that keeps improving indefinitely?
-
Note
The Problem With a Blank Prompt Box
We've given people extraordinarily capable AI tools and often left them to imagine the use cases. Perhaps discovering what is possible should be part of the product.
-
Essay
If AI Does the Doing, How Do We Learn the Judgement?
AI can increasingly produce work that once required years of accumulated skill. That creates an enormous productivity opportunity, but it also raises a harder question about how people acquire the judgement to know whether the work is good.
-
Note
Turning Things Off Is Part of Building
Technology organisations celebrate launches because new capability is visible. Decommissioning creates a quieter kind of value: less risk, less complexity and more freedom to change what comes next.
-
Note
The Boundaries Around What We Can Do Have Moved
The most interesting strategic question about AI may not be how much AI an organisation should use. It may be what has become possible now that some old constraints are moving.
-
Note
Make the Safe Path the Easy Path
If AI makes delivery dramatically faster while the functions around delivery still operate at the old speed, the constraint has not disappeared. It has moved.
-
Note
Don't Automate the Workaround
AI makes it increasingly easy to automate the work people already do. That makes it more important to understand why the work exists before making it faster.
-
Note
Capability Without Judgement
The dangerous threshold for agentic AI may not require human-like intelligence. A system that can pursue an objective persistently, use tools and take consequential actions can matter long before it understands the wider context.
-
Note
Performance Management Isn't a Favourite Rank Order
Poor performance should be addressed. But ranking everyone against everyone else is not the same thing as understanding who is performing poorly and why.
-
Note
When Do We Stop Measuring AI Adoption?
AI usage can be a useful signal while organisations build capability. It becomes a poor destination if the objective is simply to make the number go up.
-
Note
AI Cost Is an Architecture Problem
At enterprise scale, the price of the model is only part of the economics. Calls, context, routing, retries, loops, evaluations and fallbacks determine how that price becomes a system cost.
-
Note
The Scarce Resource May Be Trusted Access
As agents become capable enough to act, the limiting factor may stop being what the model can do and become how much context, permission and authority we are willing to give it.
-
Note
Building Is Getting Easier. Owning Still Matters.
AI is lowering the cost of building and experimenting with software. That makes the boundary between an experiment and an owned service more important, not less.
-
Note
What If the Most Efficient Team Isn't the Best Team?
AI may let smaller teams produce the same or greater output. But output is not the only thing teams produce.
-
Note
When the Artefact Stops Being the Evidence
AI is making polished CVs, applications and take-home work easier to produce. That may make the process behind the artefact a more useful signal than the artefact alone.
-
Note
Are We Building the Answer, or Building to Find It?
Cheaper execution does not mean we need more certainty before we build. It can mean we need less certainty about the solution and more clarity about what we are trying to learn.
-
Note
Models Commoditise. Harnesses Accumulate.
The model may be the most visible part of an enterprise AI system, but the durable capability is increasingly everything the organisation builds around it.
-
Note
Centralise the Reusable. Distribute the Problem Solving.
A specialist AI capability can be valuable, but its success may ultimately be measured by how much useful capability becomes ordinary across the rest of the organisation.
-
Note
When Writing Software Stops Being the Scarce Bit
AI is making software dramatically easier to produce. The more interesting organisational question may be what becomes scarce when engineering capacity is no longer the constraint it once was.
-
Note
Capability Without Judgement
The dangerous threshold for agentic AI may not require human-like intelligence. A system that can pursue an objective persistently, use tools and take consequential actions can matter long before it understands the wider context.