The use of Artificial intelligence (AI) across the architecture, engineering, and construction (AEC) industries is skyrocketing, with organizations eager to embrace agents, predictive analytics, and automated workflows. But as Matt Fern, National Digital Engineering Manager from Stowe Australia, highlighted during a recent Autodesk University 2026 session, many construction firms may be, “…trying to jump quickly to the end of the story,” without first addressing a more fundamental problem: they need data to support those AI ambitions.
Construction organizations certainly aren’t lacking information. They generate PDFs, spreadsheets, emails, forms, and countless other sources of project data. However, “…having lots of information does not mean you have good data,” Fern explained. Inconsistent naming, missing metadata, disconnected systems, and statuses that mean different things from one project to another all create problems that AI alone cannot solve. As Fern put it, “AI does not currently turn bad data into good data.”
Consequently, a growing disconnect exists between the industry’s adoption of AI and its progress on basic information management. Fern broke down industry research showing that while nearly half of surveyed construction businesses reported using AI and machine learning (ML), only 42 percent said site teams had access to real-time project data, while a quarter were still working from paper-based information. “AI is adopting and accelerating faster than the information environments supporting it,” expressed Fern.
Perhaps an even larger challenge is what Fern described as construction project-level thinking. Different projects can use different tools, workflows, platforms, and information structures, resulting in what he characterized as “four projects speaking four different languages.” Without a business-level approach to standardizing and retaining that information, data becomes fragmented – and organizations can be, “…genuinely good at delivering projects and still be terrible at building organizational knowledge.”
Fern then drew on Stowe Australia’s own experience tackling these challenges to identify three ways the industry can begin building the structured, consistent data foundation needed to support analytics, automation, and AI at scale.
1. Creating one data journey
For Stowe Australia, creating a stronger data foundation first meant rethinking how information moved through the lifecycle of a project. The company initially used separate templates for estimating and delivery, but separating them also separated the data.
“We eventually realized we were breaking the information journey ourselves,” Fern explained. “The same project could have information in two environments.” This created duplication, uncertainty over which information was current, and a loss of continuity as projects progressed from one phase to the next.
The lesson was that changes in internal teams or project phases should not require the information journey to start over. Data generated during estimating can remain valuable during delivery, just as information created during delivery can support handover and even future projects. Stowe Australia ultimately moved toward a, “single project environment and a continuous information journey,” where the project could change phases without its data having to “start again,” Fern stated.
“You need enough governance to protect the data, and enough agility that people don’t work around the system.” –Matt Fern
Maintaining that continuity also required greater standardization across the business. The company established a national core with baseline minimum agreed data sets, helping ensure that common elements—such as issue statuses or the purpose of particular folders—retained the same meaning from project to project. At the same time, projects could add information when their individual requirements demanded it.
“You need enough governance to protect the data, and enough agility that people don’t work around the system,” said Fern. The goal, he explained, is to “…have a national core, have a baseline of minimum agreed data sets. [And] outside [of] that, allow people to be flexible.”
By combining a continuous project environment with a consistent baseline for how information is structured, construction organizations can begin replacing fragmented project data with information that remains usable across phases, teams, and ultimately the wider business.
2. Using an information management framework
But establishing consistency across projects raises another question: what should that information structure actually look like? For Stowe Australia, answering that meant moving beyond a system inherited from its paper-based past and adopting a recognized framework for managing information.
Fern pointed to the company’s former. “One to 20 folder structure,” as an example. Despite making the transition to digital workflows, the structure itself had roots in Stowe’s paper-based system, which used physical folders purchased with tabs numbered one through 20. “We replicated that in a digital world,” Fern explained.
“…AI needs structured information at the time of recording.” –Matt Fern
Stowe ultimately turned to ISO 19650, an international standard for managing information throughout the lifecycle of built assets. For Fern, its value was less about adopting terminology and more about providing. “…a recognized information framework for something we were already trying to solve at an organization level.”
That framework helped the company consider fundamental questions about its data: What information did the organization need to know? What information needed to be created at the project level? What needed to be exchanged? And how could that information be delivered consistently?
“The point is that AI needs structured information at the time of recording,” expressed Fern. “…ISO 19650 gives us a useful framework for managing structured information or data.” Rather than beginning with an AI-specific framework, Stowe Australia’s experience demonstrated the value of first establishing an information management foundation capable of producing the structured, consistent data that future AI applications will depend on.
3. Turning structured data into intelligence with Autodesk Forma
While ISO 19650 provided Stowe Australia with a framework for managing structured information, putting those principles into practice required advanced tools capable of capturing that information consistently. Autodesk Forma, formerly Autodesk Construction Cloud (ACC), became part of that practical foundation.
“Set up your issue templates in Forma. Set up your custom fields in Assets. Set up your review structures,” Fern said. “Use them across the entire business.” Applying common templates, fields, and structures helps ensure that information captured across different projects follows the same language and can be analyzed beyond an individual job.
By getting the data foundation right today, the industry can put itself in a stronger position to take advantage of whatever AI capabilities come next.
Fern also encouraged construction organizations to understand what their existing platforms already offered before building another database, spreadsheet, or custom system. “Forma already captures a lot of structured information,” he explained, noting that some is available through out-of-the-box reporting in Insight Builder, while more can be accessed through application programming interfaces (APIs).
Ultimately, preparing construction data for AI does not begin with AI itself. Stowe’s experience demonstrates the value of establishing a continuous information journey, applying a recognized information management framework, and using platforms like Forma to consistently capture and access structured data – creating a stronger foundation for analytics, automation, and AI at scale.
As construction organizations continue exploring what AI can offer, the quality of the outcomes will depend heavily on the information behind them. By getting the data foundation right today, the industry can put itself in a stronger position to take advantage of whatever AI capabilities come next.
