For decades, digital platforms ran on a simple bargain. You typed a query, an engine returned options, and you did the rest: researching, comparing, deciding, booking. Generative AI changed the middle of that, summarising complex information on demand. The bigger shift is the one just ahead. AI is moving from answering questions to resolving intent.
Today you search for a plumber. Soon you will tell an assistant to fix your leaking tap and expect the whole thing handled: finding a checked trade, working out who is genuinely available, getting a firm price, booking the slot, paying for it. Every one of those steps is an action rather than an answer, and each has to be wired into systems that can carry it out.
As CTO of Checkatrade, my job is to make sure our architecture, trade ecosystem and operating model are built to lead in that world. That means rethinking how trust works, and using AI to deliver real efficiency on both sides of the job.
Verified trust as infrastructure
Traditional moats are eroding quickly. A year ago, proprietary job pricing data looked like a defensible advantage. Today a large language model can estimate job costs with 80% to 90% accuracy. Once foundation models reproduce static information this easily, information on its own stops being an advantage.
What holds is verified trust, and consumers are increasingly proxying that trust to AI models.
Think about how much you trusted ChatGPT two years ago, then a year ago, then today. People are handing over decisions they would once have made themselves and eventually, the consumer will stop choosing a trade from a shortlist at all. They’ll trust the assistant to choose.
That raises the stakes on whatever sits behind the recommendation. An AI agent completing a high-stakes transaction in the physical world needs verified identity, accredited vetting and genuine reviews tied to completed, audited jobs.
Our human-in-the-loop vetting and industry benchmark regulatory compliance mean that when an AI delegates a task, it connects to a fully checked professional. If a business holds core truths that no model can reproduce on its own, it has a real moat. Ours is verified, real-world trust.
Reimagining the consumer experience
Homeowners put off smaller jobs. Oven cleaning, hanging a door, clearing the gutters. Finding someone for a quick fix, chasing quotes and waiting for callbacks takes around three days of follow-ups, and plenty of those jobs get abandoned.
Moving from search to execution requires converting unstructured inquiry into instant, structured booking. This is what we found when building Checkatrade Express, for example – the primary challenge wasn’t just user interface, but creating layers capable of managing pricing, availability, and payment in real time.
Beyond simple booking, we found that genuine automation requires conversational AI layers to ask targeted questions about access materials, the age of a property – all until a job is accurately scoped.
Meeting trades where they already work
For the first time, we can build software that meets tradespeople in the way they already work, rather than forcing them to change how they work to use our software. That is the transformational unlock.
The admin burden is real: an average of two hours every evening handling paperwork, writing quotes, chasing unpaid invoices and answering customer enquiries. Complex enterprise software was never going to fix that.
The key is building ambient, asynchronous tools. In developing trade management platforms like TradeMore, our focus has shifted to AI voice assistants that handle calls on site and automated inboxes that handle customer messages, quote creation, chasing and invoicing.
This solves the latency gap between customer intent and trade execution. If a customer expects a job resolved in minutes, someone has to respond in minutes, and a trade halfway up a ladder cannot. An AI workflow maintains that real-time momentum without forcing the worker to stop their manual labour – and gives their evenings back, too.
Operational discipline and the AI workforce
Internally, our technology strategy rests on three pillars: AI-Powered People, AI-Scaled Operations and AI-Native Products. AI is not magic. It is an operational discipline that needs clear guardrails, enterprise security and practical adoption.
We are all in on AI tooling for our engineers. Working in high-trust environments with Claude Code, Cursor and Gemini, tasks that once took weeks now take days, more than doubling our engineering throughput. Across London and our growing technology hub in Barcelona, we treat AI models as adaptable commodities and keep our proprietary data, trust networks and workflow engines at the heart of the strategy.
The resilience of practical work
As AI reshapes entry-level white-collar roles, it is also showing the value of practical, hands-on trades. The UK faces an urgent skilled labour shortage: 900,000 active tradespeople today, and an additional 1.3 million needed over the next decade to meet home improvement and retrofitting demand.
Young people and parents are recognising that trade careers offer strong earning potential, stability and resistance to AI disruption. Two-thirds of young people view trades as respected pathways, and nearly half are actively considering joining. An algorithm can write code or draft text. It cannot replace a plumber fixing a burst pipe or an electrician rewiring a home.
The next decade will not belong simply to the companies that automate first, but to those that make themselves worth automating into. That is what we are building, with everything underpinned by verified human trust.
