Every general contractor has sat through this exact meeting. Someone gets back from a conference keynote swearing that AI cut some other firm's schedule in half, and by the time the claim survives three retellings down the org chart it has quietly grown into half the schedule and under budget, and somebody's cousin's GC apparently framed out a hospital in six weeks flat. Nobody in the room can ever produce the model behind the number. It is preconstruction's version of a fish story, and every estimator has learned to nod politely and get back to the actual bid due Friday.

That gap, between AI saves a ton and an actual defensible number, is exactly what Suffolk Construction and MIT set out to close this week. On September 16, Suffolk, working with the MIT Center for Real Estate and the MIT Media Lab's City Science group, released a joint study called Construction in the Age of AI, and it does something refreshingly boring for an industry addicted to big round percentages: it shows its work.

The study grounds its numbers in a sample 180,000 square foot multifamily project and traces six places where AI is already changing how that project gets built: design automation, offsite manufacturing, permitting, scheduling, skilled labor and subcontracting, and supply chain and procurement. Applied together across that sample project, Suffolk's model puts the combined savings at 17 to 20 percent on total cost and 22 to 25 percent on schedule. Those are the kind of numbers that usually show up on a keynote slide with no footnote attached. Here they come with one.

The footnote matters. Suffolk and MIT are upfront that these are modeled figures, not audited results pulled off a finished job site, built from academic literature, case studies, expert interviews, survey responses and an industry roundtable of more than 50 construction leaders, not a controlled trial against a matched project. James Scott, co-lead at the MIT Center for Real Estate, framed the study as a starting point rather than a verdict: "The next step is to keep building the data foundation needed to understand where AI has the strongest impact, where the limits still are and how those findings can be translated into better decision-making across real projects." Suffolk chairman and CEO John Fish put the stakes more plainly: "Construction is at an inflection point. AI presents a real opportunity to transform the way we build. The most meaningful progress will come when the entire ecosystem aligns around better data, smarter workflows and shared accountability."

Two of those six levers, skilled labor and subcontracting, and supply chain and procurement, sit almost exactly where BidForgeAI spends its own time: getting the right subcontractors matched to the right scope at the right price before a shovel ever moves. Seeing an MIT-vetted model name that stretch of preconstruction as one of the industry's biggest AI opportunities, rather than the flashier drone-and-robot end of construction tech, is a useful data point for anyone still deciding where their own AI budget goes first. It is not the part of the job that makes for good trade show video, but it is apparently where a fifth of the money is sitting.

It is also worth reading this study next to the rest of what has shipped this month. Procore shipped a ranked-shortlist agent for staffing. Alpha Vision pitched agents reading existing security cameras for safety compliance. OpenSpace pushed reality capture into automated owner reporting. Every one of those is a vendor's pitch for what their specific tool does. Suffolk and MIT's study is the first attempt this month at drawing the whole map at once and admitting, in writing, which parts of it are still guesswork. That is a genuinely different kind of contribution, and it is worth more to an owner sizing an AI budget than any single feature announcement, precisely because it says where the study's own confidence runs out.

None of this kills the conference-keynote fish story. Someone will still round 20 percent up to half by the time it reaches the site trailer, because that is simply what numbers do once they leave a slide deck and enter a break room. But now there is an actual paper to point at when they do, with the six levers, the sample project, and the caveats laid out for anyone willing to read past the headline figure instead of just repeating it.

Sources