The Routing Algorithm That Became a Union Organizer's Best Friend

When the Math Doesn't Add Up

A logistics optimization platform designed to squeeze more efficiency out of delivery fleets is now being used by the people doing the actual delivering — to prove their employers are asking for the impossible.

Delivery drivers at two regional carriers recently started plugging their assigned routes into their company's own AI routing software. They entered the official service windows, required break times, and posted speed limits. Then they ran the simulation.

The result: the same system their management uses to plan routes calculated they should be making 96 stops in a 10-hour shift. Their actual assignment? 118 stops. That 23% gap between what the AI considers safe and legal versus what dispatch actually assigns became evidence in a safety complaint and talking point in a union organizing drive.

The routing software is sold by a North American logistics SaaS company marketing the standard value proposition: AI-based route optimization, predictive ETAs, labor and fuel cost reduction. According to posts in driver organizing forums, workers figured out they could access the vendor's "what-if" simulation tools through standard dashboards or demo environments. Feed in the constraints. Compare the output to reality. Screenshot the difference.

The Unintended Witness

What makes this particularly awkward for employers is that the logs and reports from these AI systems are timestamped, archived, and maintained by the vendor. A B2B SaaS product designed to optimize operations has become third-party evidence that the operations might not be optimizable — at least not within legal and safety boundaries.

In internal organizing chats that spilled into public driver forums, workers shared screenshots showing management assigning 15 to 20% more stops per route than the system calculated as realistic with breaks and safety margins enabled. The same efficiency recommendations the AI generated were being cited by management when pushing higher quotas — except with the break and maximum-speed constraints quietly relaxed or removed.

One driver noted in a forum post that they were deliberately breaking up their workday to match the system's parameters, then documenting every instance where the actual route made that impossible. "The computer says I need 45 minutes of break time and can't average over 55 mph. I'm showing them I can't do both and hit my numbers."

When Optimization Tools Get Optimized Against You

This situation exposes something most B2B software vendors probably don't think about: once your tool's calculations become canonical inside an organization, they create a shared reality layer. Any gap between how the system is configured for planning purposes and how work actually gets assigned becomes discoverable. And in sectors with organized labor or active organizing campaigns, discoverable means actionable.

For logistics companies, the problem is straightforward. They bought software that promises to show the most efficient way to run routes. When workers use that same software to demonstrate that assigned routes exceed what's efficient — or safe, or legal — the vendor's own algorithm becomes the argument against current practices.

This also hints at a next phase in labor negotiations where unions might start asking for direct access to routing models, forecast algorithms, and the configuration presets used to generate schedules. If the math is already there, why shouldn't the people subject to the math be allowed to see it?

The Productivity Paradox Strikes Again

The broader pattern here is familiar to anyone watching enterprise AI deployments: tools built to optimize operations often surface the gap between how operations are supposed to work and how they actually work. Sometimes that gap is a good thing to surface. Sometimes it's uncomfortable.

In this case, a routing optimization platform designed to help companies do more with less is helping workers demonstrate that "more" might mean "too much." The software didn't pick a side. It just did the math. But when the math contradicts the mandate, someone has to reconcile the difference.

For now, that reconciliation is happening in organizing meetings, safety complaints, and screenshots passed around driver forums. The routing algorithm didn't set out to become an organizing tool. It just turns out that when you build software to calculate what's actually possible, people on both sides of the efficiency question want to know the answer.