Tracking what got built last week is still one of construction’s dumbest time sinks. Someone walks the site, marks up a plan, argues with the schedule, and three days later produces a report that’s already stale. DroneDeploy wants that job gone. Its new Progress AI reads drone flights and 360-degree walkthroughs and spits out a progress report in minutes.
How it works
The tool runs vision-language models against captured site imagery and compares what it sees to the plan, flagging what’s complete and what’s behind. DroneDeploy pegs the output at 95% accuracy and says it’s more than 100 times faster than manual tracking. During development the system ran across more than 50 projects, from commercial offices to hyperscale data centers, the kind of sprawling sites where a human progress walk can eat a full day. Progress AI is one of three agents the company unveiled at its Horizons event, alongside Safety and Inspection AI, with beta deployments of autonomous ground robots slated for later this year.
The business signal matters as much as the demo. DroneDeploy says it hit break-even and raised a strategic $15 million specifically to push the progress product. Contech funding has concentrated hard on AI this year, and a profitable vendor raising to accelerate a shipping feature reads differently than a pre-revenue startup chasing a pitch.
The honest caveat
Ninety-five percent accurate is a marketing number until a superintendent stress-tests it on a messy site, and progress tracking is exactly where the last five percent gets expensive. Occluded work, staged-but-not-installed material, and anything hidden behind finishes are the cases that fool a camera. The tool’s real value probably isn’t replacing the walk on a complex interior; it’s killing the routine exterior and structure tracking on big, repetitive footprints, which is why data centers keep showing up in these pitches.
Still, the direction is clear. Owners on projects like a high-rise office build increasingly want progress data continuously, not monthly, and the labor to produce it by hand doesn’t exist. If the accuracy holds up on live jobs, Progress AI is less a novelty than a preview of how schedule verification stops being a person’s job at all.