Three Roadblocks to Scaling AI in AECO

What separates AI success from stalled projects

The conversation around Artificial Intelligence in Architecture, Engineering, Construction and Operations (AECO) has shifted from theory to practical exploration. 84% of AECO leaders report AI positively impacts productivity. Yet the gap between successful pilot and meaningful integration remains a challenge. 

During the Expert Edge webinar about AI, three AECO experts from ARKANCE, Lantis and Autodesk share their real-world experiences and advice. Each voice provides transformation journey insights to help you identify the roadblocks and best practices to successfully scale AI. Here is what they have learned.

“The single cultural mandate every CEO must act on is this: recognize that AI is already here, and your employees are already using it. But to scale it safely, leadership must establish a clear triad: strong Leadership, strict Governance, and continuous Education.” Grégoire Arranz, Global CEO of ARKANCE

Organisations that adopt AI unlock potential for efficiency and innovation, but the pitfalls are real and only deliberate leadership can navigate them successfully. As Greg noted live, the productivity gaps in our industry are already massive. Waiting for the technology to fully mature before jumping in means risking your ability to compete at all.

Three roadblocks quietly stalling your rollout

AI projects rarely fail because of the technology itself. They fail because of what sits underneath it. Three roadblocks appear consistently across AECO organisations and all three can be identified before a single tool is selected.

ROADBLOCK 1  Fragmented data environments

Data readiness isn't a technical roadblock, it's a strategic one. Half of AECO leaders cite integrating AI with existing systems as a major challenge. Almost always, that integration problem points to the same root cause: data that isn't structured, accessible, or trustworthy enough to build on.

Emmanuel Di Giacomo, BIM & AI Expert at Autodesk, puts the sequence plainly: before an organisation can benefit from AI, it must first define its strategic objectives, put the right resources in place, assess the data available, and set achievable KPIs with a clear timeline. Not the other way around.

Organisations staying competitive are deploying collaborative common data environments (CDEs) to serve as this digital backbone, structuring and securing information before AI interacts with it. Currently, 96% of construction data goes unused because it remains trapped in departmental silos. Bringing this information into a central CDE improves trust and breaks down these barriers, but it also addresses a critical security vulnerability: an unmanaged backbone risks exposing confidential project information to the wrong people.

Gregoire warns, "This foundation is important for the AI to feed on information and provide accurate results. But it’s also critical for security. If you don't get organized properly, confidential project data could be made available to many more people than you want."

Without a strong data foundation, AI doesn't fail dramatically. It quietly underdelivers.

ROADBLOCK 2  Tool-first thinking

The teams that get past the pilot phase don't hand AI adoption to the technology department, they hand it to the people closest to the work. The most common mistake organisations make is treating AI as an IT project.

Yanissa De Jonghe, Head of Digital & Data at Lantis, is direct about where the real measure of success sits: "AI only creates value when it connects to the real work. AI is not the duty of an IT or digital department, it needs to be part of the DNA of an organization to have real impact".

That means finding internal sponsors, giving business process owners a seat at the table, and building a safe environment where testing something and failing isn't a setback — it's part of the strategy. "Create a culture where it's acceptable to test things and if you fail, fail fast and fail forward," says Yanissa.

The fastest returns often come from the least glamorous processes. AECO software platforms are embedding AI directly into the tools teams already use. While this lowers the technical barrier, Yanissa found that the fastest and most lucrative ROI at Lantis came from automating non-core processes.

High-administration tasks like flagging invoice anomalies and answering contractual questions are less digitised, which means AI has more potential for immediate impact. Assessing these tasks and defining a strategic objective is a good starting point. 

Grégoire agrees with this incremental approach, noting that breaking innovation down into "chewable bits" is the fastest way to learn and fail forward. However, he notes that quick wins must be anchored centrally: "Modularity enables speed, but a coordinated program to implement the right policies and learning practices is essential to allow capitalization of those learnings across the whole company." 

ROADBLOCK 3  Underestimating the adoption curve

Traditional assumptions about who resists new technology and who leads adoption,  are being turned upside down in AECO. Ignoring the human rhythm of adoption is a governance failure that stalls even well-funded programmes.

Yanissa has observed this directly at Lantis: the most active early adopters are experienced roles, project directors and senior assistants, who use AI to accelerate productivity without fear of replacement.

We see this same trend accelerating across the industry, particularly through Agentic AI (AI designed to autonomously execute workflows). Grégoire notes that the most transformative immediate impacts are happening in data- and context-intensive roles, specifically RFQs, quotations, accounting, and customer support. At this stage, however, the mandate isn't total task delegation; it's using AI as a human productivity accelerator.

To scale past this roadblock, true integration requires meeting employees where they are, blending specialized technical training with hands-on education to build lasting confidence in these advanced solutions. Elevating your team's skills ensures that these embedded AI platforms actively empower the people driving the project.

The question isn't ROI anymore — it's RONI

For decision-makers still on the fence, the conversation has shifted. The risk isn't in investing in AI. It's in waiting — and the risk of not investing (RONI).

While traditional ROI models are tough to calculate because the technology's history is short, the cost of inaction is clear. Greg revealed that early adopters are already capturing 5% to 10% productivity gains. In a tight-margin industry like AECO, a 10% efficiency gap creates an immediate, insurmountable competitive disadvantage for those who chose to "wait and see."

"It is a mistake to think that digital transformation starts with a tool, rather than a clear organisational need, or that AI investments require massive capital and slow ROI," says Yanissa.

She calls it the "cost of legacy": the longer an organisation delays building its data foundation and change capability, the more expensive future adaptation becomes. Lantis experienced this in real terms when COVID hit, their digital readiness allowed 300–400 external contractors to move to remote work on day one. That wasn't luck. It was the result of decisions made years earlier.

Emmanuel says the future of digital transformation lies in Outcome-Based BIM, and that embedded AI tools can help AECO organisations clean up their data pipeline to reach BIM maturity faster. When assessing solutions, he recommends ensuring they support three pillars: automate to reduce repetitive tasks, assist to surface actionable insights, and augment to enhance designer creativity.

The gap between pilots and integration isn't a technology problem — it is a data and people problem, in equal measure. The organisations that get the most from AI aren't just the ones with clean, well-governed data. They're the ones where people at every level are championing adoption, asking better questions, and building new habits.

Data gives AI something to work with. Culture gives it somewhere to go. Get both right, and the foundation you build will actually hold.

Partner for innovation

Grégoire frames the ideal relationship not as a vendor transaction, but as a true partnership: "We see this as co-invention work. Our customers own the intellectual property of their data and industry know-how—they know how to design efficiently and build right. ARKANCE brings the method, the data structure, and the experience to deploy the technology in the most safe, secure, and unlocked way."

ARKANCE partners with organisations to guide, equip, and empower teams through AI adoption, from building the right data foundation to scaling with confidence. Explore the Expert Edge series for real-world strategies and lessons from experienced leaders shaping change. Transformation takes a team. We're your people.

Contact us to start building your AI roadmap.