Will AI Fix Construction, or Just Automate Flawed Data?
By Michael Pink, CEO & Founder, SmartPM
The artificial intelligence boom has officially reached the built environment. For construction owners and developers funding massive capital projects, this technological leap sounds like the ultimate solution to the industry’s most persistent headaches. But a stark contrast exists between the promise of AI and the current reality of construction outcomes. Today, the construction industry globally loses an estimated $1.85 trillion annually to bad data. When that bad data leads to delays, the financial fallout is severe: recent industry reports show the average construction dispute in North America now exceeds $60 million and takes over a year to resolve. Furthermore, SmartPM’s own analysis reveals that over 70 percent of projects fail to meet their baseline schedule expectations.
The financial risk of ignoring these slipping schedules is massive. A reliable rule of thumb is that one day of delay typically costs at least 0.01 percent of the total project value.
If we have all of this new technology at our fingertips, why is the failure rate still so high? The answer lies in the fundamental rule of any computational system: “garbage in, garbage out.” AI is only as useful as the data it analyzes. In construction, the foundational dataset for measuring time, risk, and execution is the project schedule. Unfortunately, the data living inside those schedules is overwhelmingly flawed.
The “Garbage In” Problem
Before owners push their general contractors to implement AI-powered project delivery, they must understand the current state of construction schedule integrity.
Recent analysis of tens of thousands of construction schedules reveals a startling disconnect. A staggering 88 percent of baseline schedules fail basic quality benchmarks, meaning they lack the structural integrity needed to effectively plan a job. Despite this widespread lack of reliability, 75 percent of industry professionals state that the schedule is often or always used in executive-level decision-making.
The source of truth fails as a primary tool, but why? Why is this data so flawed? Most schedules are not wrong in dramatic ways; they are subtly unreliable, which makes them problematic decision tools when pressure hits. Schedules rarely drift because of sudden failures, but because of over-optimism. One person ignoring one plausible pitfall, then another ignores yet another “minor” issue. It adds up. As projects face pressure, teams often stop asking “what is actually happening?” and start asking “how do we keep the date?” Applying generic large language models (LLMs) to this defensive optimism will not fix the project. While advanced AI and analytics can absolutely test alternate sequencing and run what-if scenarios, doing so on a schedule lacking logic integrity simply produces guesses disguised as data.
Mandating “Schedule Truth” First
For owners and developers, the path forward requires a shift in expectations. Before relying on AI, owners must first mandate high-integrity, quantitative schedule data from their contractors to ensure these systems are learning from objective reality.
To protect their investments, owners must embed lightweight, tool-agnostic visibility standards directly into their RFPs and project kickoff meetings. Here is a practical framework owners should demand to establish schedule truth:
1. Require Machine-Readable Data
A PDF of a Gantt chart is not a schedule. Owners must mandate the submission of machine-readable baseline and update files (the native CPM files) with complete IDs, logic ties, and calendars. Clean, consistently maintained schedule data is the foundation everything else rests on. AI and analytics platforms cannot analyze flat images; they require the raw data to evaluate schedule health and predict risk.
2. Establish a Schedule Health “Gate”
Before any project dates are allowed to move in a monthly update, the schedule should be forced through a structural quality check. Owners should require contractors to run basic integrity diagnostics to scan for missing logic, negative float, excessive constraints, and overly long activities. Tracking a rigorous project health index gives owners real-time visibility to predict outcomes before finish dates actually move. If the schedule fails these basic quality benchmarks, the logic must be fixed before progress is officially recorded.
3. Maintain a Clear Change Journal
Schedules naturally evolve, but changes to durations and logic must be deliberate and transparent. Owners should require a “change journal” or a brief narrative alongside every update documenting what changed and why. This is supposed to happen anyway, but it certainly is not enforced. Treating each update as a document with evidentiary value is how owners protect their margins and avoid disputes if an issue arises later.
4. Track Near-Critical Volatility
Delays rarely happen overnight. They are preceded by tiny shifts in the schedule’s heartbeat. Owners should require visibility into the near-critical band of activities. Activities with small amounts of total float can easily become the project’s longest path with a single slip. Monitoring this volatility provides an early warning system for sequencing conflicts before they manifest as physical roadblocks on the jobsite.
The Bottom Line for Owners
When owners set the expectation that the schedule is the center of gravity, projects perform better. AI will not fix construction by simply layering algorithms over chaotic processes. It will fix construction by processing clean, high-integrity data at a scale humans cannot match. For owners, the mandate is clear: demand schedule truth first, and then let the technology optimize it. The choice for owners is simple: AI cannot fix what bad data has already broken. Those who demand clean, high-integrity schedules now will be ready to capture AI’s real value later. Everyone else will just automate the same mistakes at scale.
Michael Pink, founder and CEO of SmartPM Technologies Inc., is a construction industry professional with over 20 years of experience in project controls, risk management, and dispute resolution. As a consulting professional, Michael specialized in capturing, organizing, and analyzing complex data sets related to commercial construction projects and converting them into meaningful insights designed to positively influence the bottom line. Michael’s experience spans construction projects of all sizes and scopes across all sectors throughout the globe.
Having seen firsthand the immense value that construction data analysis offered to the built world, Michael founded SmartPM in Atlanta, Georgia, in 2016 – a technology company that delivers advanced analytics in an automated fashion to stakeholders involved in commercial construction projects. By leveraging a multitude of trade secrets, artificial intelligence, and machine learning, Michael’s approach to SmartPM is rooted in a commitment to practical problem-solving and helping project teams make informed, data-driven decisions.