The offering is the reality capture company’s first foray into the AI agent landscape.
Over the past few years since the release of OpenAI’s first release of ChatGPT, almost every industry around the world has been working to figure out how to leverage artificial intelligence to improve their workflows. That includes all forms of AI and machine learning, but generative AI such as the aforementioned ChatGPT is central. The construction industry was no exception to the AI boom, even though it was it is still not clear exactly what place generative AI will be in the sectorleaders are keen not to be left behind here, as they have likely happened with other forms of technology in the past.
We are now seeing companies from across the AEC industry begin to present new offerings based on artificial intelligence, covering all parts of the workflow. A particularly interesting example, which directly benefited from these broader developments around generative AI, is emerging DroneDeploy. Early this month, the reality capture company announced the release of their new one Safety AI solution, which comes as an add-on to the DroneDeploy Ground product.
This tool looks to solve one of the biggest problems on construction sites, which is ensuring that there is a safe environment for workers and that workers take all required precautions to keep the site safe. As they note in their press release announcing the launch, more than 167,000 lost-time injuries will be recorded in the United States alone in 2022, according to the Bureau of Labor Statistics. Safety AI takes the footage already captured during a construction project and detects all visible safety hazards according to U.S. Occupational Safety and Health Administration (OSHA) standards, ranks them by severity and automatically notifies the project’s safety team.
Philip Lorenzo, DroneDeploy’s Product Director for Artificial Intelligence, told Geo Week News that this is the company’s first AI agent and that the company was uniquely positioned to develop this type of product given all the imagery their customers have long gather on the platform.
Lorenzo, who previously founded ConstructionSitewhich was eventually acquired by DroneDeploy, has been working with AI for years and says it was only in the last twelve months that these types of tools were really possible. But right when the time came, the DroneDeploy team was working to develop these types of safety-based agents.
“The technology to create something like a Safety AI, this level of an AI agent that can look at images, was not available over 12 months ago,” Lorenzo said. “As soon as it was available, it was a sprint to get it working with our machine learning and photogrammetry models. Frankly, there are many years of machine learning and a decade of photogrammetry on which it is built.”
The biggest challenge for the DroneDeploy team was to get the model to actually identify the real problems, rather than just find assets. Lorenzo gives an example of a system that can identify five ladders on a construction site – something that doesn’t really help a project team – versus being able to identify that two of them are in a position that could cause injury.
“What people really want,” Lorenzo tells Geo Week News, “is why and where things are unsafe, much more than just where things are.”
One of the keys for the DroneDeploy team was to ensure that the information collected from Safety AI was not only valuable to the end user, but that the right people were getting the information, and at a pace that made sense. Project superintendents and safety managers are already inundated with enormous amounts of messages and emails, so a conscious effort has been made to ensure that only relevant staff receive reports, and that they arrive as a full report rather than with every possible alert.
With any product that uses AI, there is always concern about the potential for job losses as functions are automated. Lorenzo expected DroneDeploy to be able to hear safety managers’ complaints, but that was not the case. Instead, he says, “they seem undeniably to love it,” and that they have been the “greatest champions.”
Instead, the backlash came more from younger and less experienced regulators, who were initially skeptical of what appears to be another set of eyes scrutinizing their turf. Lorenzo says the fact that the reports only go to pre-selected team members and that they simply follow OSHA standards has been the answer to these potential problems.
As with any AI product, this is simply the first version of Safety AI, and Lorenzo notes that there are some future upgrades in the works. For example, today the software only includes images sourced from 360-degree cameras as most of their customers already use these tools and they provide a good view of the workplace, but adds that they will add additional sources in the future. Additionally, the reports are based on today’s OSHA – which is US-centric, of course – but will eventually add standards from international agencies for those customers as well. In the meantime, however, OSHA’s standards overlap quite a bit with those of other similar organizations around the world.
For now, DroneDeploy hopes their first foray into the realm of AI agents will result in safer standards. Given their technological knowledge and the real-world data already collected by their customers, Lorenzo says it was essential for the company to create these types of tools as soon as they were reasonably able to do so.
“As soon as OpenAI started going multi-modal – meaning they can see photos, the first thing we had to do was make the world a safer place and workplaces safer,” Lorenzo said. “It’s almost like we had a moral obligation to do that.”
