But the TCO managers also wanted to talk about something else. “We have a lot of workers in the oil fields. It would be nice to know where they are and what they are doing,” one manager said. “If they are doing anything at all.”

This is what our Chevron partners were most keen to discuss: how to better surveil their workers. TCO had thirty or forty thousand workers on site, nearly all local Kazakhstanis. They worked on rotating shifts — twelve-hour days for two weeks at a time — to keep the oil field running around the clock. And the managers wanted to use AI/ML to keep a closer eye on them.

They proposed using AI/ML to analyze the video streams from existing CCTV cameras to monitor workers throughout the oil field. In particular, they wanted to implement computer vision algorithms that could detect suspicious activity and then identify the worker engaging in that activity. (My Microsoft colleagues and I doubted the technical feasibility of this idea.) Enhancing workplace safety would be the reason for building this system, the managers claimed: more specifically, they hoped to see whether workers were drunk on site so that they could dispatch help and prevent them from hurting themselves. But in order to implement this safety measure, an “always-on” algorithmic monitoring system would have to be put in place — one that would also happen to give management a way to see whether workers were slacking off.

The TCO managers also talked about using the data from the GPS trackers that were installed on all of the trucks used to transport equipment to the oil sites. They told us that the workers were not trustworthy. Drivers would purportedly steal equipment to sell in the black market. Using GPS data, the managers wanted to build a machine learning model to identify suspicious driving activity. It’s not a coincidence that minor tweaks to the same model would also allow management to monitor drivers’ productivity: tracking how frequently they took bathroom breaks, for example, or whether they were sticking to the fastest possible routes.

The TCO managers were also interested in Microsoft products that could analyze large quantities of text. “Let’s say we have the ability to mine everyone’s emails,” one executive asked. “What information could we find?”

When I reflect back on this meeting, it was a surreal experience. Everyone present discussed the idea of building a workplace panopticon with complete normalcy. The TCO managers claimed that monitoring workers was necessary for keeping them safe, or to prevent them from stealing. But it wasn’t convincing in the slightest. We knew that they simply wanted a way to discipline their low-wage Kazakhstani workforce. We knew they wanted a way to squeeze as much work as they could from each worker.

I held my tongue and made sure to appear calm and collected. So did my colleagues. Collectively representing Microsoft, we turned a blind eye, and played along perfectly. We sympathized with TCO’s incriminating portrayal of their Kazakhstani workers and the need to uphold the rule of law. We accepted their explanation that increased surveillance would improve worker safety. But truth be told, we didn’t even need the excuses. Microsoft was hungry for their business. We were ready to concede.