What does AI customer service look like on a gig-economy platform?

On a gig-economy platform, customer service is two-sided. Workers (drivers, riders, sellers, freelancers) and end customers each have separate needs, separate urgency profiles, and separate leverage. The platform's AI has to handle both, with distinct rules, distinct tones, and distinct escalation paths. The same AI cannot use the same approach for both sides without one side feeling worse than the other.

A rider on a delivery platform reports their last earnings were short by ten dollars. An end customer reports their food arrived cold. Both are now in the same support queue. The platform's AI handles each. The contexts are different: the rider depends on earnings accuracy for their living and has time pressure (the next shift); the customer wants a refund or replacement and has reputation pressure (they will rate the platform). The AI's tone for each cannot be the same.

What people in the field are saying

The CS Cafe writes about the multi-stakeholder nature of modern support in its career-track series. Gig platforms are the most visible case: the worker is also the platform's customer in a real sense, with different rights and different vulnerabilities.

What does the worker side need?

Fast resolution of earnings disputes. Clear policy on what is paid for and what is not. A path to a human for cases that affect their livelihood. Information about platform changes that affect their work (algorithm changes, fee changes, region changes). The AI's job here is largely transactional with high stakes.

What does the end-customer side need?

Order issues resolved fast, often during the experience itself (food late, ride cancelled). Trust that the platform stands behind the gig worker's work without dismissing legitimate complaints. A path to compensation when something goes wrong. The AI's job here is closer to consumer support, with the speed expectation of real-time.

Where does the policy tension live?

Whose word counts when the worker and the customer disagree. The customer says the order was wrong; the worker says it was correct. The AI cannot adjudicate; the platform's policy has to. Most platforms tilt one way by default and tilt the other in specific cases. The AI applies the tilt at scale, which makes whatever bias the policy has visible.

Where does AI go wrong on gig platforms?

Treats workers with the same dismissive default that consumer chatbots often have, even though workers' stakes are higher. Treats end customers as second-class because the worker is closer to the platform. Applies policy without acknowledging the worker's livelihood pressure. Each erodes trust on the affected side.

What is the practical pattern?

Two separate AI configurations, one for each side of the platform, with different tone, scope, escalation thresholds, and policies. Shared customer-state where helpful (the order both sides are discussing is the same order), separated where not. Audit both sides separately; trust on one side does not generalise.