Construction is the world’s largest sector, accounting for roughly 15% of global GDP. It’s also historically been one of the least digitized and least productive. That paradox is exactly what drew Puneet Raj, CEO of Fieldwire, which is a platform for managing construction jobsites on more than four million projects worldwide, into the industry.
“I was working at Amazon on the Kindle when the Hilti Group reached out and asked me whether I wanted to help digitize the world’s largest sector,” Raj recalls. “I liked the challenge and jumped right in.”
Fieldwire, founded in 2013 and acquired by the Hilti Group in 2021, sits at the intersection of that digitization challenge. According to Raj, the key to unlocking AI in construction isn’t in the back office; it’s on the field.
The Chaos and Beauty of Field Data
When asked what field data actually looks like on a typical job site, Raj spoke freely. “It’s as chaotic and beautiful as you can imagine, since 80% of construction data is created in the field.” But it is rarely captured, and it changes almost daily.”
Field data, he explains, takes many forms: forms filled out daily, WhatsApp and text messages flying between crews and supervisors, pin-ups on walls or printouts to indicate the latest status, phone conversations between field crews and office staff, and RFIs lost in long email threads. It’s large, it’s unstructured, and that’s both the opportunity and the challenge for any AI system. However, it’s missing the critical element of context.
“If you have an issue on the field, it’s because you are at a particular specific location. It’s specific to that location, specific to the trade, and specific to the task. If you lose this context over time, the data is pretty much useless.”
A punch list discarded at the end of the day has no value. However, map that same punch list to a specific task, location, and trade, and suddenly AI can work its magic by turning unstructured information into something useful for the future.
Putting Data in Context: How Fieldwire Works
To explain how Fieldwire ties tasks, photos, and RFIs to specific plan locations in real time, Raj asks us to imagine a day in the life of a construction craftsman.
“The moment you enter a job site, you take a look at the plan and locate yourself on it. Let’s say it’s a high-rise; we’re on floor two, room 20. You can automatically see your list of tasks, checklists, and any RFIs already raised on the asset you’re working on.”
If a worker finds a clash or an issue, they can immediately take a photograph, write a description, and pinpoint it on the plan. Everything stays in context from location, task, and trade.
“It’s a living trail,” Raj says. “Even after the project is completed, you have a trail of what happened at that particular location during that job.”
That trail becomes a learning tool. An AI system that can read all of this historical data and flag that when different contractors work on the same job site simultaneously, there are no surprises. “It’s not just useful for completing this particular project on time; everything you do feeds into creating better projects in the future.”
3D Models Belong in the Field, Not Just the VDC Trailer
Historically, Virtual Design and Construction (VDC) teams sat in trailers, worked with rich 3D models, and printed out 2D plans for field workers. The problem is that the moment something is printed, it’s already outdated. “I’ve been in trailers where people ask, ‘It’s clashing in the model, but is it also clashing in the field?'” Raj recalls.
“There can be such a disconnect between what’s happening in the model and reality.”
That’s changing. With more powerful BIM viewers, including one Fieldwire is building, field teams can now navigate complex 3D models directly. This is especially critical in sectors like data center construction, where conduit and pipe density is so high that 2D plans are insufficient. “3D does belong in the field. It has to move beyond the VDC departments into the field. The industry has been ready for this change, and tools like Fieldwire are helping advance it.”
Real-World Impact: Saving Hours and Millions
When asked for a concrete example where better field-captured data changed a project outcome, Raj points to both macro-level data and a specific case study.
At the macro level, Fieldwire’s annual customer surveys have consistently shown that the platform saves up to one hour per day, per construction worker on the field. At the project level, he shares the story of a large hospital build with multiple subcontractors and a general contractor. Early in the project, a QA/QC process implemented through Fieldwire caught that certain pipes had been routed in the wrong place on day two.
“If this had been discovered the normal way, during an owner walkthrough toward the end of the project, the hospital would have been delayed by several months, and the rework cost could have gone into the millions of dollars,” Raj explains. “But because there was a standard QA/QC process in the handoff between trades, the foreman was able to catch it early and save both time and money.”
The Future: AI That Belongs on the Field
While much of the construction AI conversation focuses on back-office applications, Raj believes the real opportunity is on the job site itself.
“A lot of people are working on AI for people in the back office. But we believe AI also belongs on the field. We believe AI can help save up to three hours per craftsman per day by taking away a lot of the busy work, like auto-filling forms and automating RFIs and submittals.”
That vision is what drove the recent launch of Field Intelligence, a field-first AI platform built into Fieldwire. Its goal: integrate all field data, make it specific and contextual to the task and location, and automate the busy work so craftsmen can spend more time doing wrench work rather than paperwork.
Construction generates more data than almost any other industry, but most of it is rarely captured and quickly lost. The platforms that succeed in bringing AI to construction won’t necessarily be the ones with the most sophisticated back-office algorithms. They’ll be the ones that capture field data with the right context and turn it into a living, learning system that makes every future project better than the last.
