AI Stops Helping and Starts Running the Show

At Mobile World Congress 2026, Charles Li, President of Huawei's Chemical and Building Materials Business Unit, put it plainly: AI is no longer sitting on the edge of construction and manufacturing. It's in the engine room now.

"In recent years, most AI applications have been concentrated in auxiliary production scenarios, such as detecting whether employees are wearing safety helmets," he said. "The biggest change now is that AI is moving from a supporting role to a leading role, truly penetrating core production systems."

The practical example is Conch Cement, where Huawei and its partners have deployed an optimisation system that integrates AI models with industry production data. It analyses more than 100 real-time parameters - raw materials, process conditions, kiln chemistry - and adjusts production accordingly. Coal consumption has fallen by 1%, which sounds modest until you put a number on it: around $250,000 saved per production line, per year.

But, there is a context issue worth naming. Most of the cases highlighted at MWC are in heavy manufacturing - cement, chemicals - rather than building and infrastructure projects. The shift is real, but its translation to on-site construction in the UK is not yet straightforward. The principles are the same. The implementations are not.

The UK construction industry now has a formal benchmark for net zero carbon performance. The Net Zero Carbon Buildings Standard, developed with Bureau Veritas, provides the industry with clear performance criteria for measuring and reducing the environmental impact of buildings - addressing long-standing concerns around vague sustainability claims and greenwashing.

This is not another aspirational framework. It's designed to give clients, contractors and supply chains a consistent, verifiable measure. The global construction sector is responsible for a significant share of carbon emissions, material consumption and waste. The industry has been working towards net zero for years without a shared definition of what that actually means in practice. This standard changes that.

For contractors and project teams, the immediate question is how this integrates with existing green building certifications - BREEAM, NABERS, the Future Homes Standard - and whether clients will start requiring compliance in procurement.

Construction firms are rolling out generative AI copilots fast. The use case is compelling - search project documents, summarise emails, pull key dates from a schedule. With tight margins and chronic labour shortages, the speed gains look attractive.

But, Construction Dive has been tracking a growing concern: when these tools get it wrong, the consequences are not abstract. They land on live projects. A wrong date in a programme summary, an incorrect clause surfaced from a contract, a fabricated figure in a cost report - these aren't edge cases in a research paper, they're the kind of errors that create disputes, delays and liability.

Construction is different from most industries that have adopted AI copilots. The documents being processed are legally binding. The projects have fixed deadlines and penalty clauses. The information is often highly project-specific, which means general models trained on publicly available data don't always have the right context to interpret it correctly.

The industry is not being told to stop. The tools have genuine value. But, the question of who is responsible when an AI-generated summary leads to a bad decision is not settled - and in construction, unsettled liability questions tend to cost money.

Dreamer, the new consumer-first platform from former Stripe CTO David Singleton and former Android executive Hugo Barra, launched in beta this week. The pitch is simple: a place where anyone - technical or not - can discover, build and use AI agents. Think of it as an app store, but the apps are intelligent and they work together.

The platform's "Sidekick" agent builds agents on your behalf through natural language. You describe what you want, the Sidekick builds it, tests it and hosts it. No database configuration, no API keys, no deployment headaches.

It's early. But, the direction it points to matters for construction. The tools being built on Dreamer - calendar agents that research your meeting attendees, self-completing to-do lists, personalised briefing podcasts - are the same categories of time-saving automation that construction project managers spend their lives doing manually. The technology is not new. But, access to it without a developer is.

Tool builders on the platform get paid based on usage. Dreamer hosts everything. And the platform uses whichever AI model performs best for a given task, without the user needing to know which one that is.