AI readiness in construction: the three pillars every CTO needs
Are you prepared for what’s next in AECO?
In a recent episode of Bridging the Gap: The Conversation, CIO under pressure, Atul Khanzode, CTO at DPR Construction, shared how one of the largest general contractors in the US is approaching AI readiness. Rather than treating it as a technology problem, DPR sees it as an organizational one. With 29 years at DPR and responsibility for all technology functions at a firm operating across mission critical, life sciences, healthcare, and advanced technology sectors, his perspective is grounded in the reality of building at scale.
The AI readiness gap nobody wants to admit
Before any construction firm can get value from AI, it has to be honest about where it actually stands. Not where it aspires to be, not what the press releases say, but what the data foundations look like, how mature the internal skills are, and whether the organization has the governance in place to use AI responsibly on sensitive projects.
Most firms haven't had that conversation yet. They've run pilots, experimented with off-the-shelf tools, and watched their teams use ChatGPT informally. But experimentation without foundation doesn't scale. And the pressure to deliver on AI — from boards, from clients, from competitors — is arriving faster than most organizations can build the infrastructure to support it.
Atul Khanzode has been thinking about this longer than most. DPR invested in its data warehouse five to six years ago, specifically to build the foundation that AI would eventually require. That decision looks prescient now. At the time, it was simply disciplined.
The firms that are furthest ahead on AI readiness today are almost always the ones that made unglamorous infrastructure investments years before AI became a boardroom conversation.
The ones scrambling now are the ones that skipped that step.
Want to hear how a CTO thinks about the full technology mandate? Watch the full conversation with Atul Khanzode here.
The three pillars of AI readiness at DPR
Atul's framework for AI readiness at DPR is built around three pillars. None of them start with the AI tool itself. All of them have to be in place before the tool can deliver value at scale.
Pillar 1: Data foundations
The first and most fundamental pillar is data quality. AI models are only as good as the data they're trained on. Without clean, structured, governed data, AI outputs are unreliable at best and actively misleading at worst.
DPR's investment in its data warehouse half a decade ago means the firm enters the current AI moment with a foundation that most competitors are still building. But data quality is only part of the challenge. On the types of projects DPR works on — mission critical data centers, life sciences facilities, healthcare — the data itself is highly sensitive. Clients come with NDAs. Contracts specify exactly what can and cannot be done with project data. The governance requirements are significant.
"We are stewards of data on the projects that we work with them on. There are very strict guidelines in our contracts on what we can do with data."
Atul Khanzode, CTO, DPR Construction
The solution DPR has developed is a structured approach to data anonymization that allows AI models to be trained on project data without exposing proprietary client information.
What is a data clean room?
A clean room is a controlled data environment where raw source data is anonymized before being used to train AI models or shared across organizational boundaries. At DPR, the clean room sits within their data warehouse and acts as the bridge between sensitive project data and the AI models being built on top of it. Source data never gets shared directly. Anonymized data goes into the clean room, where it can be used to improve services and train models without violating client NDAs or data governance requirements. It's a concept more familiar in digital advertising and financial services than construction — but as AI adoption accelerates across the industry, it's likely to become standard practice for any GC working on sensitive project types.
Key takeaway: AI readiness starts with data quality, not AI tools. If your data foundations aren't in place, no AI tool will deliver consistent value. For firms working on sensitive project types, a structured approach to data governance and anonymization isn't optional — it's a prerequisite.
Pillar 2: Experimentation to scale
The second pillar is the mechanism for moving from isolated experiments to scaled adoption. Most construction firms have pockets of AI experimentation happening informally. The challenge is turning those experiments into repeatable, scalable workflows that deliver value across the organization.
DPR's approach is to work directly with operations teams and functional work groups to identify pain points and quick wins — specific, high-friction tasks where AI can demonstrate clear value fast. Contract review is one example Atul cites. Once a use case is proven, the goal is to scale it quickly so that enough people see the value to drive broader adoption.
"We are working with our operations teams as well as our other functional work groups to identify what are their pain points and what can be quick wins that they can adopt for AI. We're trying to take that and scale it up quickly so that people see the value."
Atul Khanzode, CTO, DPR Construction
This bottom-up approach — starting with real problems in specific functions rather than top-down mandates — is one of the most reliable paths to AI adoption that sticks. When the use case is genuinely useful to the people doing the work, adoption follows naturally. When it's mandated without connecting to a real pain point, it doesn't.
DPR also maintains an innovation team specifically tasked with managing this experimentation process — evaluating what works, what doesn't, and what's worth scaling. The discipline of that process matters as much as the experimentation itself.
Takeaway: Scaling AI adoption requires a structured process for identifying pain points, proving value quickly, and moving successful pilots into standard workflows. Innovation teams without that process produce experiments. Innovation teams with it produce capability.
Pillar 3: Internal education and training
The third pillar is the one most firms underinvest in, and the one that determines whether the first two pillars actually deliver value. Data foundations and experimentation infrastructure mean nothing if the people who need to use the tools don't know how, or don't trust them.
DPR has made internal education a significant organizational commitment. Of the firm's 6,000 administrative employees, approximately 2,500 have already been trained on AI fundamentals — how to use the tools, how to construct effective prompts, and how to apply AI to their specific role. That's not a pilot. That's a program.
"Unless you do that, people don't really adopt some of these tools, or they might use it once in a while, and then they don't really do anything with it. So we are emphasizing quite a bit this internal training and education component."
Atul Khanzode, CTO, DPR Construction
The training is role-specific rather than generic. Understanding that different functions require different AI workflows — and building training around those specific workflows rather than generic AI literacy — is what separates programs that change behavior from ones that produce a one-day workshop and nothing more.
The next phase at DPR goes further still: working with functional teams to identify specific focus areas where AI can make measurable progress in the work they actually do.
Takeaway: AI training that works is role-specific, practical, and connected to real workflows. Generic AI literacy programs produce awareness. Role-specific training produces adoption. The firms that invest in the latter will compound that advantage over time.
The four buckets: how to evaluate any AI investment
Alongside the three pillars, Atul offers a framework for evaluating whether any AI investment is worth making — one that cuts through the noise of feature lists and vendor promises and gets to what actually matters for a construction business.
Every technology investment, AI or otherwise, should be evaluated against four questions:
- Does it improve revenue and margin?
- Does it reduce the cost of doing business?
- Does it improve the employee experience?
- Does it improve the customer experience?
If a tool can answer yes to one or more of those questions with evidence, it earns its place. If it can't, it doesn't matter how impressive the demo was.
"Instead of getting enamored by just the technology tools, really having a dialogue internally about what is it that you're trying to accomplish — that will guide your further technology exploration decisions."
Atul Khanzode, CTO, DPR Construction
This is the same principle applied throughout DPR's technology strategy more broadly. Software spend gets rationalized against how well it advances the firm's core strategy — delivering predictable outcomes through VDC, prefabrication, and self-performed work. If a tool advances that strategy and the results are visible, it stays. If it doesn't, the process takes care of itself.
Takeaway: Before evaluating any AI tool, define what success looks like in terms your business already uses. Revenue, cost, employee experience, customer experience. If the tool can't move one of those dials, it probably isn't worth the investment regardless of what the headlines say.
Where to start if you haven't started yet
The honest message from Atul's experience at DPR is that the firms getting AI right didn't start with AI. They started with data. They built the warehouse, focused on quality, established governance, and created the conditions for AI to work before AI became a priority.
For firms that haven't taken those steps yet, the starting point isn't the most advanced AI tool on the market. It's an honest assessment of data quality, a conversation with operational teams about real pain points, and an investment in the education infrastructure that will determine whether any AI tool gets genuinely adopted or quietly abandoned after the pilot.
If your coordination data isn't connected, structured, or accessible, that's where to start. Talk to Revizto about building the foundations your AI strategy needs.
For teams looking to build digital skills and get more from their coordination workflows, the Revizto Academy offers free, self-guided courses built around real construction workflows — no client account required.
FAQs
AI readiness in construction refers to the organizational, data, and skills foundations a firm needs to have in place before AI tools can deliver consistent value. This includes clean, structured data, governance frameworks for sensitive project data, internal training programs, and a structured process for moving from experimentation to scaled adoption. Most firms have experimented with AI but fewer have built the foundations required for it to work at scale.
AI models require clean, structured, and well-governed data to produce reliable outputs. For construction firms, this means investing in data warehousing and data quality before deploying AI tools. Firms working on sensitive project types — mission critical, life sciences, healthcare — also need data governance frameworks that allow AI to be trained on project data without violating client NDAs or data sovereignty requirements.
A data clean room is a controlled environment where raw project data is anonymized before being used to train AI models or shared across organizational boundaries. It allows construction firms to use project data as a training resource for AI without exposing proprietary client information. The concept is more established in financial services and digital advertising but is becoming increasingly relevant for GCs working on sensitive project types as AI adoption accelerates.
The most reliable approach is to start with real pain points in specific operational functions, prove value quickly with targeted use cases, and then scale those workflows across the organization. Working directly with operations teams to identify quick wins — rather than imposing top-down mandates — produces adoption that sticks. An internal innovation team with a structured evaluation process helps separate the experiments worth scaling from the ones worth stopping.
Generic AI literacy programs produce awareness but rarely change behavior. The most effective AI training is role-specific, connected to the actual workflows of the people being trained, and followed by function-specific focus areas where AI can make measurable progress. DPR Construction has already trained approximately 2,500 of its 6,000 administrative employees on AI fundamentals, with role-specific training and prompt engineering as part of the program.
Evaluate every AI investment against four questions: does it improve revenue and margin, does it reduce the cost of doing business, does it improve the employee experience, and does it improve the customer experience? If a tool can move one or more of those dials with evidence, it earns its place. If it can't, the sophistication of the technology is irrelevant. Starting with a clear definition of what success looks like for your specific business is more valuable than any feature comparison.
Generative AI has the potential to transform how construction firms handle document-heavy, data-intensive workflows — from contract review and specification analysis to design optimization and coordination. For general contractors, the most immediate applications are in tasks that currently consume significant manual time: summarizing meeting minutes, producing first drafts of documents, searching across large volumes of project information, and analyzing data patterns across projects. The firms getting real value from generative AI in construction right now are the ones that started with clean data foundations, identified specific high-friction use cases, and invested in role-specific training before deploying the tools at scale.
