Stop Chasing AI: Build the Data Backbone That Actually Improves Profitability

After 30 years of broken promises, one industry veteran believes the real revolution isn't about technology. It's about actually joining the dots.

Here's a question that cuts to the heart of why construction remains stubbornly resistant to change: if we've been predicting digital transformation in this industry for three decades, why are companies still passing tender documents around in physical envelopes?

That observation comes from Iain Curtis, a chartered civil engineer who has spent four decades watching construction attempts, has largely failed to evolve. He witnessed the dot-com boom of the late 1990s, worked inside a pioneering online tendering platform when 56K dial-up modems were the cutting edge of internet connectivity, and has now spent 15 years building data analytics software specifically designed for construction businesses. His conclusion? The industry hasn't fundamentally changed since he started his career in 1982, despite a relentless stream of new technologies that promise to revolutionise everything.

Yet Curtis isn't pessimistic. He's frustrated. The productive kind of frustration that comes from seeing a problem clearly and knowing exactly what needs to happen to solve it. His company, Single Point Solutions, tackles something that sounds mundane but proves transformative: helping construction firms understand how their actual business runs, not just what's happening on site. This subtle yet crucial distinction explains why many digital initiatives in the construction industry produce impressive dashboards yet fail to significantly impact profitability or efficiency.

The Silo Problem Nobody Fixed

Step back to the late 1990s. Curtis was working for a major contractor, watching how the business actually operated. What he found was not unusual; it was structural. Different departments operated almost as separate businesses within the same company:

• The commercial team sat at one end of the office, negotiating contracts and managing client relationships

• The estimating department occupied its own isolated corner, running calculations in what Curtis describes as "their black hole"

• Project management existed separately, focused on site delivery and operational logistics

These departments barely communicated except once a month when the CVR (Cost Value Reconciliation report) landed on desks like a postmortem examination of the previous month's performance. This wasn't deliberate dysfunction. It was simply how construction had organised itself.

"The whole of the industry, yes, costs a value reconciliation report, and it's understanding the profitability really on a month-by-month basis of where your project is," Curtis explains. "So with the CVR, that's the Bible really."

Here's the structural problem: by the time you're examining the CVR data, the money is already spent. You're looking backwards, not forwards. The silos mean that nobody truly understands in real-time whether your business is actually making money or heading toward a loss. Individual departments optimise for their own metrics; estimating for speed, commercial for margin, and project management for delivery, without coordinating toward actual profitability.

Nearly three decades later, Curtis's clients face virtually identical challenges. Departments still operate semi-independently. Integration remains fragmented. Real-time visibility remains a theoretical ambition rather than an operational reality. The technology has improved dramatically, but the fundamental architecture of how construction companies share information and make decisions has barely evolved.

What's remarkable isn't that the problem persists. What's remarkable is that solving it requires technology specifically designed for construction's particular dysfunctions, rather than generic business software. Curtis built Single Point Solutions around the concept of a "construction data layer" - essentially a translation system that sits between enterprise resource planning (ERP) systems and analytics tools, standardising data so it actually makes sense across departments.

"We've got a curated layer," he notes. "That's part of our main IP and our product base."

Without that curation layer, connecting artificial intelligence directly to messy, decades-old database structures simply doesn't work. Yet that's precisely what many technology vendors are now promising to do.

The AI Hype Versus Reality Gap

Here's where the conversation gets uncomfortable. We're in another technology boom. The parallels to the dot-com era are unmistakable, but so are the differences. During the late 1990s, Curtis's company was building a genuinely valuable online platform for tenders, but the infrastructure to deliver it simply didn't exist. ADSL hadn't been invented. Broadband was science fiction. The technology was ahead of the market.

With artificial intelligence, the market is clearly ready. Construction firms desperately want solutions. The problem now is the opposite: vendors are making promises that the technology can't yet keep.

"The hyperbole and saying 'what AI can do' and 'what the reality is' are two different things at the moment," Curtis observes. "Technology companies are saying, 'Yeah, we're AI-driven, we've got AI this.' And you scratch beside the server, and it's very simple. It's just basically ticking a box."

The danger becomes apparent when you examine how firms are rushing to deploy AI without understanding what it actually requires. A vendor might propose placing an AI agent over your data and connecting it to your systems. It sounds elegant, sounds automated, sounds transformative. Then comes the technical reality check: What's the state of the data?

If you've got 20 years of accumulated database entries with inconsistent formatting, missing values, and structural problems that nobody documented, strapping AI on top is like giving a Formula One car to someone without a driving licence. It won't work. The technology will either produce garbage outputs or simply fail.

What Curtis identifies as the critical preconditions for effective AI deployment:

• Data quality must be genuinely good and well-structured, not just available

• Historical data from legacy systems often requires complete cleanup before AI can be applied

• Consistency across decades of records is essential; inconsistent formatting and missing fields sabotage algorithms

• Governance frameworks need to be in place before deployment, not after problems emerge

"If it's an old database structure of a system somebody's had for 20 years, masses of data, can you stick AI over it? Absolutely not. You've got to sort the data out," Curtis explains. This isn't pessimism about AI's potential. It's clarity about what happens when you skip the foundational work.

The concern deepens when you consider data security. Recent research from leading AI safety organisations reveals that rogue data or prompt injections embedded within documents can corrupt AI outputs, potentially spreading harmful or maladaptive information throughout institutions. Poor quality data is problematic enough. Genuinely compromised data deployed through AI systems becomes an institutional risk.

Why the Small Contractors Are Winning

Here's an unexpected insight from Curtis's experience: Smaller contractors are often more agile about adoption than their larger counterparts. The conventional wisdom suggests that scale brings advantages. More resources, more sophisticated IT departments, more budget. Yet Curtis finds the opposite pattern.

Larger contractors, despite their advantages, tend toward more cautious technology adoption. They've got legacy systems they've sunk millions into maintaining. They've got established processes that, while inefficient, are at least understood. Their IT departments focus on keeping the lights on rather than transformation.

Smaller contractors operate differently. They're still adaptable. They lack entrenched systems. They recognise that agility might be their competitive advantage and treat technology adoption as an existential necessity rather than an optional upgrade. When they embrace analytics and data-driven decision-making, they experience impact quickly.

"The size and scale of a business should not be a barrier to good BI, good analytics and good AI, potentially as well, because it should be there for the masses. It's not just for the big boys," Curtis emphasises.

This observation flips a common narrative. In construction, conventional wisdom suggests that sophisticated technology belongs to large organisations with dedicated IT resources. Yet Curtis's experience suggests that smaller firms, precisely because they're less constrained by legacy infrastructure and established processes, can leapfrog larger competitors by embracing integrated analytics early.

The Long Journey Nobody Mentions

Curtis offers an important reframing for how construction firms should think about digital transformation. It's not a project with an end date. It's not something you "complete" and then move on.

"Analytics and looking at data, it's a journey and you've got an endpoint, but actually it could be quite secure until that endpoint can change. It's to be agile. It's not to have a data strategy if they've got one, for instance, it's not set in stone and everything. It's an agile tool to live and breathe," he explains.

This challenges how many organisations approach digital initiatives. They invest in a comprehensive three-year transformation programme, create detailed project plans, expect specific measurable outcomes, and then declare victory when the defined outcomes are achieved. Then they wonder why adoption stalls and performance gains evaporate within 18 months.

The reality is messier. Your data needs are going to evolve as the business evolves. As new opportunities present themselves, your technology stack must change accordingly. Your understanding of what constitutes "good" analytics will deepen as you mature. An agile approach involves continuously refining, regularly evaluating whether your strategy still serves the business's emerging needs, and staying willing to change course. This model proves more effective than ceremonial strategic planning, which typically characterises construction business transformation.

Some of Curtis's clients have been on this journey for years, gradually building sophistication in how they operate their businesses. Yet many construction firms haven't even started. They're still organising themselves around monthly CVR reporting, still siloing departments, and making real-time decisions based on historical data.

The Real Opportunity

What becomes clear through Curtis's experience is that the real revolution in construction won't come from deploying the cleverest algorithms or most sophisticated AI. It will come from building the foundational infrastructure. Clean data, integrated systems, departments that actually communicate. That makes sophisticated tools useful.

This is where Single Point Solutions sits. They've spent approximately 10 years developing their construction data layer, the curated system that sits between legacy ERP systems and analytics tools. Only with that foundation in place does deploying AI for real decision support become possible. Only then can firms actually understand profitability in real time, anticipate cash flow problems before they become crises, and optimise supply chain management based on genuine business intelligence rather than departmental assumptions.

The opportunity for firms that get this right is genuinely enormous. They'll move faster than competitors. They'll make more accurate business decisions. They'll identify problems sooner. They'll profit from seeing the business as a whole, not in isolated departmental slices.

But it requires patience. It requires accepting that transformation takes years, not quarters. It requires investing in data quality before rushing to deploy advanced algorithms. It requires being willing to look honestly at how your organisation actually operates, then systematically improving those operations rather than bolting new technology onto broken processes.

Listen to the Full Conversation

This blog captures the broad strokes of a fascinating conversation but misses the granular details where the real insights live. In the full episode, Curtis explores construction's specific data challenges in depth, discusses how his platform is already helping dozens of firms break out of the silo problem, and debates whether the current AI boom will end in a bust. The question of whether construction could be different this time around is particularly compelling.

There's also discussion about how his 20-year experience building platforms specifically for construction has taught him what the industry actually needs versus what technology vendors think it wants. And notably, Curtis reflects on his experiences during the dot-com boom, drawing surprisingly relevant parallels to where we are now with artificial intelligence. Why the conditions this time round might actually support genuine, sustainable transformation if the industry gets the foundations right proves especially interesting.

The conversation runs roughly 45 minutes and touches on practical implementation challenges, the very real security risks of deploying AI on compromised data, and why Curtis remains optimistic about construction's potential despite his well-earned frustration with its resistance to change. If you're wrestling with how to make sense of AI's potential for your organisation or struggling with the gap between transformation ambitions and actual operational reality, this episode is worth your time.

Listen to the full Project Flux Spotlight episode with Iain Curtis to hear him explore the decade-long journey of building data infrastructure for construction, why smaller contractors often lead larger ones on this journey, and how the industry might finally break free from the silos that have constrained it for 30 years.