Insights
On AI agents that run real processes, the knowledge they depend on, and technology leadership, for mid-market and PE-backed companies. From someone who has built them.
AI agents that run your processes
What always-on agents can now do for a mid-market business, how to put one on a real process, and what keeps it safe and affordable: claims intake, credit control, the controls and the cost. See what agents can run
The knowledge layer
An agent is only as good as what it knows about your business. Why AI is held back by its records rather than the model, and how to fix that: the knowledge audit, the entity registry, retrieval and the company brain.
The capacity for greatness
We all have a capacity for greatness, and the ceiling is higher than we think. A series on personal responsibility at work: how it grows outwards in circles, why leaders are measured with a different stick, and why it matters more, not less, in the age of AI. The philosophy behind it
Always-on AI agents are here. The agent is now the easy part
OpenAI, Microsoft and Meta all launched always-on AI agents in September 2026. For small businesses and the mid-market the agent is now cheap and easy. What decides whether it helps is your records, your rules and who checks the work.
Claims intake can run itself. Claims decisions should not
How an AI agent runs insurance claims intake for MGAs, TPAs and loss adjusters, why a named handler still decides every claim, and what the FCA expects.
Most late payments can be recovered. Credit control is a job for an AI agent
Most late B2B payments can be recovered: the invoices are stuck, not refused. How an AI agent runs credit control, cuts DSO and leaves judgement to a person.
Headless is the new mobile-first. Agents are about to become your biggest user
Box, Zapier, Salesforce and Stripe all said the same thing in April: agents will become the dominant users of software. If your product can only be used through a UI, it cannot be used by them. Here's what to do about it.
AI is not just for engineering. Every function in your business can use it today
Finance, marketing, sales, legal, customer support, project management. AI tools can improve every business function right now. Here's what actually works in each, and what doesn't yet.
Traditional moats are dissolving. Size and capital no longer protect you
Scale, capital, and headcount used to guarantee dominance. AI is changing the equation. Competitive advantage now flows to the companies that move fastest and adapt smartest, regardless of size.
Agentic AI in 2026: what actually works in production
Agentic AI is moving from slides to production. The teams getting it right share specific patterns around scope, tooling, and human oversight. Here's what I've learned building agents that actually ship.
From AI that talks to AI that works: the autonomous leap of 2026
2025 gave us AI that talks. 2026 is the year of AI that works: autonomous agents that pursue your goals around the clock. Here is what is happening, starting with an open source project that grew out of nowhere.
AI is about to split the job market in two
AI will not just automate jobs. It will rebuild the entire market that matches people to work: personal career agents, employers who hunt talent before roles exist, and a bifurcation that rewards the top tier. Here is what is coming, and where you want to be standing.
The people making the biggest AI decisions understand it the least
Understanding of where AI is heading concentrates at the frontier, while the budgets and decisions sit far from anyone who has built with the tools. Why that gap matters, and how to question the advice you are given.
How to unlock AI ROI: what the 20% do differently
80% of AI projects return little because the return was never engineered in. It's fixable. The projects that deliver share specific, repeatable patterns in how they are scoped, measured, and governed from day one.
Why your AI spend isn't showing up in the numbers
Most AI spend isn't showing up on the P&L yet, but some organisations are seeing transformative returns. The difference isn't the technology. It's the deployment approach, and it's fixable.
Turning AI theatre into AI that moves the numbers
Innovation labs and pilot programmes look impressive in quarterly updates, but theatre is a choice, not a fate. Here's how to move from performative adoption to real, measurable production impact.
What to do when your AI pilot fails
Most AI pilots fail, and most failures are recoverable. Here is how to diagnose what actually went wrong, decide whether to salvage or restart, and why the people who built it are usually the wrong ones to fix it.
Stop counting AI use cases. Redesign three value streams instead
Pilot purgatory has a sequel: use-case purgatory. 47 deployed AI assistants is not a strategy. Here's how to pick three value streams and rebuild them end-to-end, the way the operating model needs to change.
Most companies are adopting AI. Few are adopting it well
84% of developers now use AI tools. A handful of organisations are getting transformative results. The difference isn't the technology. It's how it's deployed. Here's what the best teams are doing differently.
The dual-stream strategy. Protect your current business while building what replaces it
Every profitable company faces the same question: how do you invest in AI without cannibalising your own product? The answer is a dual-stream approach that protects today's revenue while building tomorrow's advantage.
If your AI strategy is just ChatGPT, you don't have one
Most leaders now equate AI with large language models. But the frontier models are available to everyone at the same price, and they are often the wrong tool. The real advantage is not the model. It is everything you wrap around it.
You don't need to build a brewery to drink a pint of beer
The most expensive software decision a founder makes is often the first one: deciding to build. Here is how to choose between building, buying, no-code, and an agency, and how AI changed the answer in 2026.
Is your software AI-resistant? How to tell if your product survives the next model
The threat to your SaaS isn't another vendor, it's how thin the product was to begin with. Four tests for whether your product is AI-resistant, and why now is the best time to build deep.
Your AI does not need a bigger model. It needs to know your business
Model capability is no longer what holds business AI back. What is missing is a structured, current, trustworthy picture of your own business: a knowledge layer. What it is, why it matters now, and where to start.
What I look at in a knowledge audit
Before any AI build, I audit a company's knowledge, not its tech stack. The five kinds of mess I look for, what a good audit should hand you at the end, and how to buy one without paying for a report that sits on a shelf.
What is an entity registry, and why does your AI need one?
An entity registry is one list of the real-world things your business deals with, such as customers, suppliers and contracts, each with a single ID and every name it goes by. A plain-English explainer, and why AI makes it urgent.
Why your RAG system keeps missing what's in your documents
Your AI assistant has read all your documents, yet it misses facts, quotes superseded policies and gives different answers to the same question. Four reasons retrieval-augmented generation falls short, and the fix.
GraphRAG vs vector RAG. Which one does your business need?
GraphRAG vs vector RAG, compared in plain English: what each is, when vector search is enough, when you need a knowledge graph, what the evidence says, and the hybrid most businesses end up with.
Email is your most underrated data source
Your systems of record hold what someone remembered to type in. Your email holds what actually happened: the commitments, decisions and exceptions. Why AI projects skip it, how to turn it into trustworthy knowledge, and how to do it lawfully.
Company brain products are useful. Most are not a brain yet
Company brain tools are good at finding documents and knowing where knowledge lives. Most stop short of reasoning across your entities, relationships and systems. What they do well, where they stop, and how to buy one without overpaying for a promise.
Agent sprawl is the new shadow IT. Your business needs a control plane
Microsoft, Google, Salesforce and IBM all shipped agent governance platforms within weeks of each other in spring 2026. Here's why every business will need an agent control plane by mid-2027, and what one actually does.
Why grounded AI agents are cheaper to run
AI is now billed by use, not by seat, and token prices are falling while AI budgets overrun. Why the cost of an agent is really a knowledge problem, where the tokens go, and how grounding and deterministic tools cut the bill.
Uber burned through its token budget by April. Your business will be next
AI FinOps is following cloud FinOps by about a decade, and the curve is steeper. Token spend is becoming a board issue. Here's why your CFO needs a token budget now, and what one looks like.
When your best AI model can vanish overnight
In June 2026 a US government directive forced Anthropic to disable its two best models worldwide, days after launch. The lesson outlasts the news: single-vendor AI dependence is now a live continuity risk. How to build optionality and model portability into your stack.
Shadow AI is your next audit finding
78% of AI users bring their own AI tools to work. That's not a problem. It's a signal of demand. Channel it with the right governance framework and you turn a compliance risk into a competitive advantage.
Who is accountable for AI on your board?
In most SMEs, no one can say who is accountable when an AI system causes harm or cost. As regulation tightens in 2026, that gap is a real liability. The fix is not a Chief AI Officer. It is a named owner and proportionate governance.
You're not talking to an LLM. You're talking to a system
When you use ChatGPT or Claude you are not talking to a model. You are talking to a system: guardrails, routing, caching, retrieval, compaction, and more. Why that matters for the AI decisions you make.
What two hours with Anthropic's agent team taught me about building AI
Anthropic ran a workshop on their Agent SDK. Five principles emerged that match everything I've seen building production agents, and one that changed how I'll build them next.
You're not a 10x engineer. You're an orchestrator, and that's harder
The best developers aren't writing code faster. They're directing AI agents, making architectural judgements, and maintaining the context that machines can't hold. The role hasn't been eliminated. It's evolved into something more valuable.
One person and a fleet of AI agents now build like a whole team
After twenty-five years climbing away from the keyboard, I went back to building. With agentic AI, the core skill has moved from recall to judgement.
75% of Google's new code is AI-generated. So what?
AI-generated code percentages are vanity metrics. They tell you nothing about whether engineering is delivering more value. Here are the five metrics your board should actually be asking for in 2026.
AI made developers 19% slower. Here's what they were doing wrong
The most rigorous study on AI coding tools found a surprising result. But the real insight isn't the headline. It's that the gap between using AI and using AI well is enormous, and closeable.
The quality you can't see is the one that kills you
Software has a visible quality and an invisible one. The visible quality wins users. The invisible quality, scale, security, maintainability, decides whether your company survives. Three war stories, and why AI makes this more urgent than ever.
Human in the loop is still the best AI model. But only when you design it
Human plus AI beats either alone, but only when the collaboration is designed. The trial where doctors plus GPT-4 gained nothing was a skills gap.
After 17 years as a CTO, I learned the engineering was the easy part
Deep technical skill is real and rare, but for a natural technologist it is the most masterable part of the job. The hard part, the part that decides whether the engineering matters, is leadership. A personal reckoning.
Agile is now the bottleneck
Agile's ceremonies were built for a world where building was slow. Agentic coding made building fast, so the heavy process now holds you back. What works instead: set a destination and navigate to it daily.
The best technology rarely wins. The best-led team does
Superior products lose to inferior ones all the time. The deciding factor is leadership, not the technology or the capital. Here is what strong leadership actually consists of, and why it drives product and market success.
How big tech wins the talent war, and it isn't the money
I spent twelve weeks interviewing for senior roles at five big tech companies to study their cultures from the inside. The thing that keeps the best people is not compensation. It is psychological safety, and it is a hard operating requirement, not a soft one.
The training ladder is broken. And nobody has a plan to fix it
AI is reshaping the junior developer role, and with it, how the next generation of senior engineers gets built. The companies solving this now will have the strongest engineering teams in five years.
If I were a junior developer today, I would use AI to outpace the broken ladder
Junior hiring is paused and everyone is betting on seniors. But the same AI that broke the old apprenticeship hands juniors a faster one, if they build.
Your world-class engineering team is already in the building
Most leaders who inherit an underperforming technical organisation reach for the wrecking ball: reorganise, replace the stack, replace the people. It is almost always the expensive mistake. The better approach is to sculpt the organisation you already have into the one you need, without ever stopping delivery.
Why your tech recruiter isn't on your side
Recruiters are paid when people move, not when they stay, so a stable team is bad for their business. Here is how the model really works, the fees, the tricks, and how to use an agency without getting fleeced.
The most important skill in business and life
Many skills matter for success. But the one that separates the best leaders from the rest, especially in the age of AI, is the ability to make good decisions when nobody knows what happens next.
The most powerful yet overlooked tool in a leader's arsenal
Gallup tracked 3,400 employees over two years. Those who received quality recognition were 45% less likely to leave. The tool that produced this result costs nothing and takes ten seconds.
Your ceiling is higher than you think
I was the overweight kid picked last for every team. At 50, I am the fittest I have ever been. Why your ceiling is higher than you think, and what keeps it low.
Your job description is the smallest circle
Responsibility grows outwards in circles: your tasks, your team, your company, the world. How far out you take it, through ownership not hours, decides your impact.
The minimum is a habit too
Every working day you choose between doing the minimum and pushing a little further. Both become habits, and over years they compound. It is not about hours.
How to build an AI strategy without a CTO
You can build a sound AI strategy without a CTO, and without being technical. What you cannot do is hand the question to whoever is selling you the technology. Here is the method for non-technical founders and SME leaders.
Does your business need a CTO?
The answer is almost always yes, but not for the reasons most people think. The right technology leadership pays for itself many times over. Here's how to think about it.
Fractional CTO. Is outsourced technology leadership right for you?
Fractional CTO engagements give you senior technology leadership without the full-time cost. Here's how to structure one that actually delivers.
Fractional CTO or AI consultancy? Don't hire the consultancy first
An AI consultancy delivers a project and leaves. A fractional CTO owns the outcome. They do opposite jobs, and hiring them in the wrong order is how AI projects become expensive, unowned messes. The honest UK answer.
The hidden conflict of interest in hiring a fractional CTO
A fractional CTO supplied by an agency has two employers: you, and the firm that pays them. That is a conflict of interest. Here is how to spot it, how to spot a fake CTO, and how to hire technology leadership that actually serves you.
No articles match your search.
Which process would you hand to an agent first?
Tell me what you're working on. I'll respond personally. If there's a fit, we'll take it from there.