Interviews with 15 Amazon Flex drivers reveal how algorithms hire, rate, and fire contract drivers with little or no human oversight
Stephen Normandin spent almost four years racing around Phoenix delivering packages as a contract driver for Amazon.com Inc. Then one day, he received an automated email.
Context & Ripple Effects
Amazon Flex has been under a slow-burn documentation arc since 2017, when reporters first embedded with the Uber-like app coordinating its self-employed last-mile drivers. The economics kept deteriorating on paper: a 2018 study found contractors earning about $11/hour after expenses despite the $15/hour headline raise, an Atlantic ride-along came in under minimum wage after costs, and a 2019 investigation caught Amazon at times using driver tips to backfill promised base pay.
The new reporting closes the loop on who controls the job itself: interviews with 15 drivers, including Phoenix veteran Stephen Normandin, show algorithms hiring, rating, and terminating contractors — Normandin's four-year run ended with an automated email — leaving no human in the loop to appeal to.
First-order effects
- Amazon Flex contractors like Normandin can lose their livelihood via automated email, with no manager or HR channel to contest the rating or reinstatement decision.
Second-order effects
- The termination mechanism stacks on top of the program's already-documented pay erosion — the $15/hour raise drivers never actually received and the tip-skimming practice the LA Times exposed — giving labor advocates and regulators a cumulative case file rather than isolated complaints.
Third-order effects
- If algorithm-only lifecycle management holds across last-mile fleets, contract delivery converges on a structure where the platform's software is simultaneously employer, evaluator, and court — pushing worker-classification disputes toward whether an algorithm can legally 'fire' someone it technically never employed.
The trend: Gig logistics is moving from app-coordinated piecework to fully algorithm-managed worker lifecycles, where hire-to-fire decisions happen with zero human intervention.