By Satish Thiagarajan, founder of Brysa, a Salesforce and data consultancy based in the UK. His company advises media, industrial, and services clients on using Data Cloud and Agentforce to turn signals into action. His work focuses on closing the loop between insight and execution in sales, marketing, and service.

Construction has no shortage of enthusiasm for AI. 56% of construction organisations are planning increased AI investment in 2026, and the business case is clear enough – top contractors solving their data integration challenges are seeing project cost reductions of 10-15%, budget and timeline deviations cut by 10-20%, and productivity gains through reductions in engineering hours of up to 30%. Yet 45% of construction firms report no AI implementation at all, with only 1.5% using it across multiple processes and fully embedded, organisation-wide AI use reported by less than 1% of participants.

The gap between intent and outcome has a straightforward explanation, and it isn’t the quality of the tools.

Construction businesses rarely grew with data architecture in mind. A project management platform added here, a finance system there, separate tools for estimating, procurement, site operations, and customer records. Each one introduced to solve a specific problem at a specific moment. Rarely connected. Construction workers currently spend 18% of their time simply searching for information, with 43% believing better data access would directly improve their decision-making. That’s before AI enters the picture.

AI works by identifying patterns across data. It depends on relationships between events, not just the events themselves. When project data sits in a BIM model, financial data lives in an ERP, and field updates come in through a separate mobile tool that doesn’t sync in real time, those relationships are difficult to establish. A model may still produce outputs, but those outputs are built on incomplete or conflicting information. The result isn’t transformation — it’s a more expensive version of the same problem.

The timing issue compounds this. Much of the data construction businesses rely on is updated in batches or reconciled after the fact. Cost reports produced weekly. Progress updates logged at the end of a shift. Variations processed days after they occur. AI built on that data embeds that delay, which limits its ability to support decisions that need to be made on site, in the moment a variation is raised or a programme slips.

The RICS AI in Construction report identified data quality and availability as a top barrier for 30% of construction firms, noting that many still struggle with fragmented, incomplete or inconsistent data — a barrier particularly acute in small and mid-sized organisations with limited digital infrastructure. System integration concerns ranked even higher, cited by the largest share of respondents, reflecting how clearly the industry understands that the problem isn’t the AI itself.

This is where “before AI” matters. In construction, that means digitising the processes still running on paper or spreadsheets, connecting the systems that hold project, financial, and operational data, and establishing a single version of project truth before asking AI to interpret it. Without that, AI adds another layer of complexity to an environment that already has too much.

Where construction businesses have done that groundwork, AI has a clear view of how projects are running and can produce outputs that reflect it — flagging cost risk before it becomes a variation, identifying programme slippage before it compounds, surfacing resourcing gaps before they hit delivery. Fragmented AI deployment, by contrast, creates data silos and limits learning loops. A single-site pilot doesn’t generate defensible advantage. Cross-project data aggregation does.

The practical implications aren’t complicated, but they do require sequencing. Connected systems come before AI-powered insights. Clean, real-time data comes before predictive analytics. A single source of project truth comes before automated decision support. Skipping those steps doesn’t stop AI from being deployed — it just changes what it can deliver.

The measurable improvements AI delivers in construction are real, but they accrue to the businesses that treated data infrastructure as the first investment, not the afterthought. As adoption increases across the industry, the gap between those firms and the ones still running disconnected systems will widen. The question isn’t whether to adopt AI. It’s whether the systems it depends on are in a state that allows it to do anything useful.