Isha Juneja, co-founder of ContraVault AI.
To anyone on a project today, it’s pretty clear construction doesn't have an information problem. We've never had more drawings, more specifications or more data. But what we do have is a decision problem, and every additional byte is making it worse.
That’s the pain that Isha Juneja, co-founder of ContraVault AI, saw first hand. So much so, that she set out to solve it. Through her time at Microsoft, Isha developed the unique ability to take complex, overwhelming information and turn it into something a team can actually act on. And yet, for someone who co-founded an AI company, she is remarkably clear that the machine doesn't get the final word.
The human stays in the loop.
A Thousand Pages and a Small Team
Many of us have been there, a large bid package landing on the estimating team's desk. Thousands of pages of drawing sheets, technical specifications, commercial terms, payment schedules and guarantees. Meanwhile, we’re a small team with a hard deadline, and a go/no-go decision that could make or break the entire company's P&L.
That’s the pattern Isha saw over and over when she moved from the tech world into construction. Teams spent a week just figuring out what was in a project before they could begin to evaluate it. Many weren't even using the simplest tools, like the search function in a PDF reader. The information was all there. It simply arrived faster than any human team could consume it.
Enter analysis paralysis. That's when abundance turns into an inability to make a decision on anything. Just because you have more information doesn't mean you make better decisions, especially when nobody has time to separate what matters from what doesn't. Instead it produces late nights, missed bids and last-second guesses disguised as gut feelings.
Reading Isn't Deciding
So, as AI became mainstream, a new way to sift through the noise has been gaining popularity. But that is also exactly where a dangerous temptation has begun to creep in.
If a machine can read a thousand pages in an hour, why not let it tell us what to do?
Isha's answer cuts right to the heart of this idea. The AI reads a bid package the same way for every contractor who uploads it. Your competitor gets the same read you do. Reading, by itself, is a commodity.
What isn't a commodity is how your company decides.
Every contractor carries its own intelligence: the risk it will tolerate, the terms it will walk away from, the lessons it paid for on past jobs. For example, Isha references a payment terms clause. One company is only comfortable with a project if it receives a 10% advance payment. Another doesn't care at all. Same clause, same document, completely different decision.
So the real differentiation comes from encoding that decision-making layer, the company's own rules and judgment, on top of the reading. The machine surfaces. The company decides.
That is the line between reading a document and understanding what matters and why.
The 80/20 Path to Scaling Capacity
In practical terms, the goal is for AI to handle the first cut, scanning the drawings, pulling the scope, flagging the risks and producing a first-pass estimate. Then the human steps in, vetting every quantity, validating the dollar value, negotiating with the customer and owning the final submittal. It’s the age-old 80/20 rule at play, only this time you’re subbing out 80% of the work to an AI.
But that extra 20%? That isn't a concession to nervous executives that don’t trust the tech. That’s where the risk lives (and construction is risky by nature), both financially and physically. The companies adopting well aren't handing over blind faith. They're handing over the tedious reading so their people can spend more time on the important decisions.
The adoption itself follows the same principle. Isha's team rolls solutions out in phases, a new module each quarter, with checkpoints along the way so people can build trust in the tool before they're asked to depend on it. Most leaders already know change is coming. What they need is permission to take it one step at a time.
And this is what changes capacity at scale. A company bidding 100 projects a year can bid 120 with the same team. A company bidding ten massive projects can go deeper on each one, with more confidence that nothing slipped through. Either way, the time that comes back isn't wasted. It goes to the human-intensive work that actually wins jobs.
In an industry that is already strapped for people, this isn't about doing more with less. It's about scaling with the people you already have.
Plugging the Leak
We talk constantly about construction's productivity problem and decades without meaningful gains. But a lot of that problem isn't about the work itself. It's about attention. Highly capable people are spending their brainpower searching, copying, checking, and chasing information instead of using it.
The leaders Isha sees making the best technology decisions don't start with the technology. They start with a question: where is our time leaking? Why didn't those two bids go out last month? What slowed us down? They stay close to their teams and keep that feedback loop running, and that's how they discover 15 people buried in manual reconciliation against a seven-day deadline that was never realistic.
Starting there keeps AI from becoming technology for technology's sake. The pain comes first. The tool comes second.
And if you have no idea where to start, Isha is quick to point out two tasks worth testing right away. The first is reconciling document sets: what the drawings show versus what the specifications require versus what the scope of work actually asks for. The second is leveling vendor quotes. Anyone who has built a bid comparison form knows the exclusions are where it gets painful, sorting out who covered what and where the gaps are.
Both tasks are repetitive, document-heavy, and exactly the kind of reading a machine can take off a team's plate so the team can focus on the call. Time how long the task takes you, then time how long it takes the AI. Note where it gets it right, where it fails, and where a human still needs to step in.
That's data, and data is what turns "we should try AI" into a business case leadership can act on.
The Decision Is the Work
At the end of the day, construction will keep generating more information. More models, more specs, more data from every corner of the job. That trend isn't going away any time soon, and really it shouldn't. But volume was never the goal.
The goal is, and always was, a better decision. A decision made faster, by the people who own the outcome.
This is simply a reminder of where the value in the data actually sits. Let the machine read. Keep the human deciding. Construction doesn't need more information. It needs more decisions it can stand behind.
Construction is cool, tell your friends!