"We're basically a computer vision-powered predictive analytics platform," says Saurabh Ladha, CEO and co-founder of Doxel, a Stanford graduate who had been interested in the construction process ever since he was a child when cost overruns on a manufacturing facility his father owned threatened to push his parents into bankruptcy. He and CTO Robin Singh co-founded the Redwood City, Calif., company in late 2015. "The fundamental issue with traditional construction software management tools or project controls platforms is the quality and the timeliness of the data because they're entirely dependent on manual entry and manual data entry sometimes comes out in a garbage, in garbage out situation. Your decisions can only be as good as the information you're getting."

Doxel's AI is informed by what it calls a Construction Encyclopedia, with hundreds of completed projects and lessons learned documented in the encyclopedia database. The AI attempts to digitize domain expertise for project controls across myriad project types. Since the AI is able to draw information from multiple different project types in the encyclopedia — such as oil and gas, health care and others—it can analyze and inform a contractor's plans based on building types and past schedules that are similar.

Ladha explained how one customer, with a large oil and gas project, using Doxel's AI figured out that over 12% of the piping in a smaller part of their facility had not yet been installed. That would be a relatively small discrepancy given that this was on a 1 million sq ft facility, but Doxel's project controls platform automatically figured out, based on current production rates of linear ft per day, that the contractor would not finish its activity on schedule and that would cause a cascade effect on the project's critical path.

"Our Construction Encyclopedia cross-references similar systems and correlates components to their relationship with cash and schedule," Ladha says. "And it's constantly learning project controls, domain expertise. We've now tracked tens of billions of dollars of construction."

One owner that is using Doxel on a major project is Shell Petrochemical Pennsylvania. Shell Petrochemical and contractor raSmith are nearing completion of what's estimated to be a $6 billion to $10 billion Pennsylvania Petrochemicals Project in Beaver County, Pa., northwest of Pittsburgh. The project includes four ehtylene processing units (three of which are polyethelene reactors), one ethane cracker, a natural gas power plant to support both the plant and the local electric grid, a 900-ft-tall cooling tower, a rail system with over 3,000 rail freight cars, numerous loading facilities for both trains and trucks. The scope of work also includes 810,000 sq ft of Ohio River dock being built by Alberici, a water treatment plant, an office building, a laboratory, and an innovation center.

"The challenge you kind of always have to balance is you have, right now, a few thousand people inside," says Dmitry Gurevich, IT director and chief information officer for the Pennsylvania Chemicals Project. "So, how do you best apply the limited resources you have from a project oversight, project management perspective that you can keep track of the things that matter the most. That's where technology liked Doxel helps tremendously, because it keeps you focused and it keeps your limited resources applied in the most effective, most efficient manner."

Shell is estimating a 2022 opening for the Pennsylvania Petrochemicals Plant, which is expected to produce 1.6 million tons of polyethylene each year once in service.

Ladha says that Doxel is focused on solving problems with the construction process that can be recognized by its AI independent of human error.

"That's the problem that we are looking to address [with] computer vision, machine vision" Ladha says. "Instead of people having to enter the data, people can focus on solving the [construction] process rather than on finding [errors] and reporting on them. It's wild."