The Algorithm Is Already the Boss: ILO Surveys Cambodia's Gig Drivers

79% of the 1,237 platform drivers the ILO surveyed in Phnom Penh said their customer rating has a very significant impact on what they earn. Not a manager. Not a wage agreement. A star average computed by an app. If you want to know what algorithmic management looks like when it is fully in charge of a workforce, Cambodia's ride-hailing and delivery sector is no longer a preview — it is the finished product.

The finding comes from the ILO's Diagnostic review of working conditions and social security coverage of digital platform workers in Cambodia, published on 2 September 2026, covering taxi, tuk-tuk and delivery drivers in Phnom Penh.

What the ILO actually measured

[Fact] The study combined 1,237 structured surveys — 415 delivery riders, 417 tuk-tuk drivers and 405 taxi drivers, sampled across all 13 districts of Phnom Penh — with 29 in-depth interviews plus key-informant interviews with platform operators, government officials and worker associations.

[Fact] The headline working conditions are stark: respondents worked an average of 6.7 to 6.8 days per week and 11–12 hours per day, including an average of 2.3 hours of unpaid waiting time. Gross monthly earnings ran from $581 for taxi drivers to $501 for delivery riders and $477 for tuk-tuk drivers.

[Fact] Context matters here: 88.3% of total employment in Cambodia is informal. Platform work did not break into a formal labour market; it grew inside an informal one.

Half the workforce has been penalized by a machine

The chapter of the report that deserves the most attention carries the quiet title "Use of automated systems".

[Fact] The platforms' algorithms directly control task allocation, ratings and penalties — and workers feel it. Beyond the 79% who tied ratings to earnings, 49% reported being penalized for refusing or cancelling orders. Taxi drivers were penalized most often (55%), followed by delivery riders (49%) and tuk-tuk drivers (44%).

[Fact] And here is the detail that should change how you read "account deactivation" stories: outright suspension was the rare penalty. Only 16% of delivery riders, 14% of tuk-tuk drivers and 5% of taxi drivers had ever had their account suspended, typically for 2.4 to 3.3 days. The dominant penalty was subtler — fewer bookings. Among penalized workers, 71% of delivery riders, 80% of tuk-tuk drivers and 82% of taxi drivers said the punishment arrived as reduced job allocation.

[Estimate] Run those two numbers against each other and the design of the system becomes visible: the invisible penalty — reduced allocation, reported by 71–82% of penalized workers — is experienced somewhere between five and sixteen times more widely than the visible one, suspension, which only 5–16% of workers have ever faced. A driver who is being punished may never be told so; a quiet stretch of no orders looks exactly like a slow afternoon. That comparison is our own arithmetic, not the report's, but both inputs sit in its figures 11 and 12.

[Fact] Where suspensions did happen, the most common triggers were customer complaints (48%) and a high cancellation rate (40%).

Platform operators did not dispute the mechanics. [Claim] In the report's interviews they confirmed that their algorithms deliberately prioritize top-rated drivers for jobs, and that low-rated drivers are contacted for retraining or guidance.

The pay figure the app shows is not the pay you keep

Gross earnings put taxi drivers on top at $581 a month. Net earnings invert the entire ranking.

[Fact] After deducting fuel, phone, equipment, registration, insurance and repairs, delivery riders kept $355 a month, tuk-tuk drivers $318 — and taxi drivers just $204, the lowest of the three despite the highest gross. The report adds that vehicle loan payments and depreciation are not included, so even these figures are on the optimistic side.

[Estimate] The report publishes gross hourly earnings ($1.90 for taxi, $1.65 for delivery, $1.50 for tuk-tuk) but stops short of net hourly figures. We did that arithmetic: at 6.7 days a week and 11–12 hours a day, a working month is roughly 330 hours. That puts net earnings near $1.07 an hour for delivery riders, $0.95 for tuk-tuk drivers and about $0.61 for taxi drivers — roughly a third of the gross taxi rate the platform dashboard displays.

That gap is not a rounding error. It is the difference between what the app tells a worker he earned and what his household actually receives.

More social security than the national average — and still exposed

Here is the counterintuitive finding, and it cuts against the standard "gig work strips away protections" narrative.

[Fact] About half of surveyed workers were enrolled in social security — 61% of taxi drivers, 53% of delivery riders and 49% of tuk-tuk drivers. Nationally, only 20.8% of Cambodia's population was covered by at least one form of social protection in 2022.

[Estimate] By our reading, that makes the surveyed platform workers roughly 2.4 times more likely to hold some coverage than the national baseline. In a labour market that is 88.3% informal, the platform is not the coverage gap — it is, oddly, a partial on-ramp to coverage.

But the fine print undoes much of the comfort. [Fact] Most of that enrolment is the voluntary National Social Security Fund scheme, which covers healthcare only — not pensions and not work injury. And work injury is precisely the risk this workforce carries: one third of delivery riders reported being injured while working. Accident insurance, where platforms provide it, applies only while a task is active; a rider hurt while waiting between orders, or while riding to a pickup, is on his own.

[Fact] Almost all respondents said they would personally bear the costs of an accident or medical emergency.

So the paradox holds in both directions: platform work delivers more paper coverage than Cambodia's informal economy at large, and simultaneously leaves this workforce's single largest occupational risk uninsured.

If you drive for a living, this is your preview

This site tracks how AI changes occupations, and most of that evidence takes the form of exposure scores and adoption surveys. Cambodia's platform sector is a different kind of data point: a place where automated systems already perform, end to end, functions that used to belong to dispatchers, supervisors and payroll clerks — the working environment of taxi drivers, delivery drivers and couriers and messengers.

The preparation lessons travel better than the wage levels do.

Track net, not gross. The starkest information gap in the report sits between gross and net pay. Whatever the dashboard celebrates, the number that matters comes after fuel, repairs, loans and depreciation.

Treat your rating as infrastructure. When 79% of workers say ratings drive earnings and operators confirm the algorithm favours top-rated drivers, maintaining a rating is not vanity — it is capital maintenance.

Assume penalties are silent. If reduced allocation is the dominant sanction, keep your own log of orders received and refused. A personal record is the only way to tell punishment apart from slow demand.

Close the injury gap yourself where schemes allow. [Claim] The report's tripartite consultations point toward extending work-injury coverage to platform workers, and interviewed stakeholders — including the platform operators themselves — agreed that regulation is needed on algorithmic transparency, minimum earnings, working time limits and safety standards. Until that arrives, voluntary healthcare enrolment plus whatever accident cover a platform offers is the fallback.

Where this evidence stops

The survey covers Phnom Penh only, and the ILO is explicit that its results are not nationally representative. Earnings and hours are self-reported. Women make up 1–8% of each worker group, too few for reliable gender estimates. And a diagnostic review describes conditions; it cannot establish whether algorithmic management caused the long hours and thin margins, or whether Cambodia's informal transport economy would have produced them anyway. [Claim] Notably, what the workers themselves asked for in interviews was not the algorithm's removal — it was higher fares, lower commission fees and access to social protection.

The machine boss is not coming for these 1,237 workers. It clocked them in years ago. The open question, in Phnom Penh as everywhere else, is whether the rules written for human bosses will be rewritten fast enough to bind the algorithmic ones.

Sources

• ILO, Diagnostic review of working conditions and social security coverage of digital platform workers in Cambodia, 2 September 2026.
• ILO, news release accompanying the report, 2 September 2026.

This article was produced with AI-assisted analysis. All figures were verified against the full text of the ILO report (74 pages); calculations labelled [Estimate] are our own derivations from the report's published data.