Shiva Dhawan, Co-founder and CEO of Attentive.ai – Interview Series

Shiva Dhawan, Co-founder and CEO of Attentive.ai, is an entrepreneur focused on applying artificial intelligence to transform infrastructure and construction workflows. Before launching Attentive.ai, he held leadership and operational roles across technology and business functions, helping shape the company’s vision around automating traditionally manual processes in industries such as construction, mapping, and geospatial analysis. Under his leadership, the company has expanded internationally while developing AI systems designed to improve efficiency in estimating, takeoffs, and infrastructure management for enterprises and contractors.

Attentive.ai is an AI-powered construction technology company focused on automating preconstruction and infrastructure workflows using computer vision and geospatial intelligence. Its platform helps contractors, landscaping companies, and infrastructure operators accelerate estimating, measurement, and site analysis tasks that have historically relied on manual labor. The company’s Beam AI product is designed to leverage aerial imagery and AI to generate highly detailed property measurements and landscaping insights, helping businesses improve bidding accuracy, reduce operational bottlenecks, and scale projects more efficiently through automation.

You founded Attentive.ai after scaling a services business in mapping and insurance, and later introduced Beam AI as your flagship product. What specific insights from that earlier phase led you to build Beam AI, and why did you choose takeoff and estimating as the entry point for transforming construction workflows?

My Co-Founder, Rishabjit and I came to the U.S. construction market during COVID, when contractors had to estimate jobs without being on-site. What kept coming up was the same constraint: contractors losing work not because they couldn’t do the job, but because they ran out of time to price it. One estimator, hundreds of pages of plans, 4 to 8 hours per job. You can’t grow a business on that.

We chose takeoffs because they’re the starting point for everything. Nothing else moves until someone measures the scope. And the output is verifiable; you either got the quantities right, or you didn’t. A 2% miss on a $10 million job is $200,000 gone. That’s not abstract. It’s a real cost that estimators carry every day.

Construction and field services are often seen as slower to adopt new technologies. What has been the biggest barrier to AI adoption in this sector, and how are you overcoming it?

Trust. Estimators have built their careers on accuracy. When they miss something, their company pays for it. So when we showed up with AI, the natural reaction was: how do I know this is right?

We didn’t try to talk people out of that concern. We addressed it directly. Every done-for-you takeoff gets reviewed by a trained person before it goes back to the customer. The automation handles volume and speed.

The QA catches anything that needs a second look. After a few jobs, customers see the pattern: the quantities are right, their team isn’t buried in plan sets, and the bids are going out faster. One of our customers, Bommarito Construction, submitted 50 more bids in six months using the platform. That’s more convincing than any demo.

Beam AI focuses on automating takeoffs, a traditionally manual and time-intensive process. Why is this workflow such a critical entry point for AI-driven transformation?

Every project starts here. Before you can price anything, someone has to sit with the plans and measure everything. One takeoff can take a full day. When things get busy, that becomes the ceiling on how much work a team can chase.

Contractors aren’t turning down jobs because they don’t want them. They’re turning them down because there isn’t time to price them.

Takeoffs also have a clear, checkable output: material quantities. You know if something was missed. That makes it a reasonable place to build trust in a new system, especially when the stakes are high.

Your platform enables companies to increase bid volume without adding headcount. How do you see this reshaping competition and margins across the industry?

It’s already happening. When a contractor can pursue three times as many jobs with the same team, they become selective. They go after higher-margin work. They can respond quickly when a big opportunity comes in, instead of passing it because they’re already maxed out.

The contractors who aren’t thinking about this are going to feel the pressure from the ones who are. Rays Stairs doubled their bid volume and grew revenue from $900K to $2 million in two months. Guardian Roofing cut takeoff time from 25 hours a week to 5. Those aren’t small gains. They change what a business can actually go after.

Beam AI incorporates a human-in-the-loop quality assurance (QA) layer alongside automation. How do you determine the right balance between AI autonomy and human oversight?

We think about it in terms of confidence and what’s at stake. The AI handles the structured, repetitive work well: reading plan sets, identifying components, pulling quantities. But construction drawings aren’t uniform. Specs can be unclear. A plan set might be missing a sheet.

The QA layer is there for those situations. For the done-for-you service, a trained reviewer looks at every output before it ships. For the automated 10-minute takeoffs, we’ve accumulated enough data, particularly in HVAC and plumbing, to move faster without that step. Steel is now rolling out soon. The level of autonomy tracks the trade and the complexity of the job.

As models improve, do you see that the QA layer becoming less central over time, or will it remain a permanent part of high-stakes workflows like estimating?

Both, depending on how you define it. The form it takes will change. A lot of what a human reviewer catches today will shift to automated checks inside the system as the models improve and we build up more data. But I don’t think you ever take verification out of a workflow this high-stakes. If a contractor is pricing a $50 million steel job, they’re going to want a checkpoint.

What we’re working toward is making that checkpoint faster and less labor-intensive. The goal isn’t to eliminate QA. It’s to make it lighter.

Attentive.ai blends AI automation with real-world operational workflows. Do you see the future of AI in construction as inherently hybrid rather than fully autonomous?

For the foreseeable future, yes. And I’d push back on the idea that “hybrid” is a consolation prize. Construction involves judgment that isn’t captured in a plan set. A good estimator knows their local subcontractor market. They know how a particular GC writes specs. They know what a job will actually cost to build, which isn’t always what the drawings say.

AI handles the quantifiable work. The human brings the context. The goal isn’t to replace estimators. It’s to get them out of the repetitive measuring so they can spend time on the work that actually requires their judgment. Which is also why we’ve built Beam AI to be an augmentator, like a plug-and-play junior estimator that handles mechanical tasks.

You’ve described AI as becoming the operating backbone of preconstruction. What does that vision look like over the next five years?

Right now we’re focused on the front end: plans to material quantities, as fast and accurately as possible. The next layer is bid management. We’ve already shipped Bid Dashboard and Bid Sniper, which give contractors a single view of their pipeline, deadlines, RFIs, and addenda.

Over the next five years, I want the platform to connect takeoffs directly into pricing and procurement. A contractor uploads plans and, within hours, has a real picture of what the job costs and what they need to source. That’s a genuinely different way of running preconstruction than what most teams are doing today.

Beam AI now supports multiple trades, from landscaping to civil and electrical work. How do you balance building generalized AI systems with the need for deep domain-specific optimization?

It’s a real tension. The underlying work is shared across trades: reading documents, parsing drawings, and extracting quantities. But the outputs are trade-specific in ways that matter a lot. How you measure HVAC equipment is nothing like measuring structural steel or concrete rebar.

We’ve built trade-specific models and invested in training data for each one. That’s why we started with HVAC and mechanical, where our data set was strongest, before expanding to plumbing and steel. We cover 15 or more trades, but we’re honest that not every trade is at the same level of maturity. We build depth as we expand.

AI is beginning to reshape traditionally offline industries. Do you believe construction could become one of the most transformed sectors over the next decade, and what would that transformation look like in practice?

I do. Part of why it’s underestimated is that it’s been so manual for so long. There’s no deeply entrenched software layer to displace, the way there is in finance or healthcare. The data hasn’t been digitized. The workflows aren’t standardized. That sounds like a problem, but from where we sit, it’s an opening. We’re not replacing an existing system. In a lot of cases, we’re building the first one.

Add to that the capital going into data centers, manufacturing, and infrastructure right now, and the pressure to price and build faster is only going up. The contractors who figure this out are going to pull ahead. The ones who don’t are going to wonder what happened.

Thank you for the great interview, readers who wish to learn more should visit Attentive.ai or Beam AI.