Dubai's Roads and Transport Authority (RTA) announced on 8 June 2026 that AI-powered dashboards at its Bus Operations Control Centre have produced a reduction of more than 68% in early bus departures. The release gives no baseline and no period, so the figure shows a direction rather than a benchmark. The lesson for any operator is simpler and more useful: early departures can be measured from AVM data, and a network that measures them can start to manage them.
What did RTA actually announce about early bus departures?
In its press release of 8 June 2026, titled "AI Contributes 68% to Reducing Early Bus Departures", RTA describes "AI-powered smart dashboards at the Bus Operations Control Centre". It says the centre "serves as the backbone of Dubai's bus network". The system was built by RTA's Public Transport Agency with the Artificial Intelligence Centre at Corporate Technology Support Services, and it runs on RTA's Dataiku platform.
The release names four applications: early-departure performance analysis, AI-powered trip cancellation prediction, rapid-response systems for metro service disruptions, and optimisation of short-turning operations. It also reports a reduction of more than 13,000 tonnes of CO2 emissions, which RTA attributes to the sustainability dashboards through less engine idling and better bus parking management.
What the release does not give is just as important: it gives no baseline or time period for the 68% figure. Those are the first questions to ask of any headline improvement, and the answers are not in the primary source.
Why do early departures matter as much as late ones?
Lateness gets most of the attention because passengers feel it directly. Early departures do the damage more quietly. A passenger who arrives at the stop on time for the published departure finds the bus already gone, and the next one may be a long wait away on a low-frequency route. The same passenger sees a delay in the app as a warning, but an early departure arrives with no warning at all.
Some regulators count both as failures. Under the contracts of Ireland's National Transport Authority, according to its Public Service Obligation Contracts Annual Performance Report 2024, a departure on a low-frequency route counts as on time only if the bus leaves no more than one minute early and no more than 5 minutes 59 seconds late. The window is lopsided: one minute of earliness is tolerated, nearly six minutes of lateness is tolerated, and everything outside that window counts as not on time. We explain the wider measurement regime in our article on bus contract penalties in Ireland.
For an operator, the consequence is that a punctuality figure which looks only at lateness can hide a real problem. If the early departures are not counted, they are not managed.
What else did RTA report in 2025 and 2026?
The 8 June 2026 release is part of a longer sequence. On 29 September 2025, RTA published a release announcing a bus punctuality improvement of more than 50%, attributed to big data, AI and advanced analytics. In it, Marwan Al Zarooni, Director of Buses at RTA's Public Transport Agency, credits real-time re-engineering of operating schedules and data-driven procedures that stop buses idling. The same release reports that bus trip cancellations fell by 4%, which RTA attributes to an AI model.
The 2025 release also describes a data-driven weather response mechanism built into RTA's crisis and disaster management framework, intended to improve passenger safety in severe weather. And it sets out a 2025 to 2030 roadmap: expanded predictive analytics, AI-supported driver training, AI-based customer service and an open data platform developed in partnership with academic and industry sectors.
All the figures in both releases are RTA's own reports, not independent audits, and they should be read that way.
How should an operator read an improvement figure that has no baseline?
Treat it as a prompt for measurement, not as a target. A reduction of more than 68% in early departures means something very different if early departures were 10% of trips, or 2%. Without the baseline, an operator cannot translate the claim into its own network.
A better use of the announcement is to ask the same questions of your own data.
- What share of departures at timing points leave early, using the threshold in your contract or your own standard?
- Which routes, times of day and drivers account for most early departures?
- How many of those departures were caught while the bus was still at the stop, and how many were found afterwards in a report?
- What is the running time allowed in the timetable on the routes where early departures cluster?
The last question is often the most revealing. Early departures are sometimes a driver-behaviour problem and sometimes a timetable problem, where the allowed running time is longer than the trip needs and the driver is simply catching up with the clock. The data can separate the two, but only if it is recorded at the level of the individual trip.
What does an AVM need to record to detect early departures?
Detecting early departures requires less than the headline suggests. The core is a record of the actual departure time at each timing point, set against the scheduled time, for every trip. On top of that, the control room needs to see the deviation while the bus is still in service, not only in next-day reports.
According to the platform page, Pysae Operations lists a live map showing delays, advances and regularity per line, calculated automatically. Advances matter here, because an advance is the measure of a bus running ahead of its timetable. The platform page also lists alerts and incidents for service breaks and critical delays, and the Pysae Driver app with a digital duty sheet, GPS guidance and two-way communication. The module page lists instant messaging between drivers and controllers. Pysae Fleet is listed as generating punctuality, mileage and journey time reports automatically from field data. For the full list of modules, see the Pysae platform.
We make no claim that these modules reproduce RTA's dashboards. RTA built its own analytics on its own platform. The point is narrower: the underlying record, a timestamped departure for every trip, is the same, and it is the foundation on which any dashboard, in-house or not, has to be built.
Which operating practices usually reduce early departures?
The RTA release does not describe the operating practices behind its result, so this section sets out general levers rather than what Dubai did. Operators commonly work on four things.
- Timing points that matter. Not every stop needs a departure check. A short list of well-chosen timing points, where the bus is expected to wait if it is early, keeps the measure meaningful and the workload low.
- Feedback to drivers. A driver who can see, on the duty sheet or through a message from the control room, that the bus is ahead of schedule can correct it before the next stop.
- Timetable realism. If the same trips run early day after day, the schedule is probably padded. Revising running times often removes the cause rather than the symptom.
- A weekly review by route. Early departures cluster. Looking at the five routes with the highest share each week is faster than reading a network-wide average.
None of these needs artificial intelligence to start. They need a reliable AVM record, a threshold agreed in advance and someone who owns the review. AI can then help to find patterns across a large network, which is the direction RTA describes.
What should Gulf operators take from Dubai's approach?
Three things stand out, and none depends on the size of Dubai's network.
First, RTA puts the control centre at the centre of the network: it calls the Bus Operations Control Centre the backbone of Dubai's bus network.
Second, the improvements are described in terms of operating measures that passengers experience: early departures, punctuality and cancellations. The environmental figure, more than 13,000 tonnes of CO2, is tied to idling and parking, which are also operating matters.
Third, RTA's 2025 to 2030 roadmap includes an open data platform. Open standards are a practical foundation for that kind of ambition. The Pysae platform page lists an API-first approach with open standards (GTFS, GTFS-RT, NeTEx, SIRI, ITxPT), and, for Operations, automatic GTFS-RT, SIRI and ITxPT exports. For the reasons real-time accuracy builds passenger trust, see our article on why accurate bus times matter, and for what separates genuine live tracking from fast refreshing, what true real-time AVL looks like.
Frequently asked questions
What did Dubai's RTA say about early bus departures?
In a press release dated 8 June 2026, RTA reported a reduction of more than 68% in early bus departures, attributed to AI-powered smart dashboards at its Bus Operations Control Centre. The release does not state a baseline or a period.
Is an early departure counted as a punctuality failure?
It depends on the contract. Under the Irish NTA's definition for low-frequency routes, a departure is on time only if the bus leaves no more than one minute early and no more than 5 minutes 59 seconds late, so an early departure beyond one minute counts as a failure.
What other results has RTA reported for its buses?
RTA reported a bus punctuality index improvement of more than 50% and a 4% fall in trip cancellations in a release dated 29 September 2025, and a reduction of more than 13,000 tonnes of CO2 emissions in its release of 8 June 2026. The 2026 release gives no baseline or period for its figures, and all of them are RTA's own reports.
Do operators need AI to reduce early departures?
Not to start. The first requirement is a reliable record of the actual departure time at each timing point, compared with the schedule for every trip. A threshold, a weekly review and feedback to drivers can all be run on that record.
Want to see departures against timetable on your own network?
If you would like to see how advances, delays and regularity appear per line in one platform, use the "Request a demo" button on this page and tell us which timing points matter on your routes.