Fleet maintenance is a universal challenge. Vehicles still need to be identified, serviced, repaired, approved, invoiced and kept compliant, regardless of whether the fleet is operating in Australia, Thailand, Japan, Europe or South America.
At Fleet APAC 2026 in Bangkok, Eden Shirley, Founder and Managing Director of Fleet Guru AI, used a live keynote demonstration to show how artificial intelligence could help fleet management companies and large operators manage maintenance across complex, multilingual markets.
The theme of the presentation was simple: Fleet is a universal language. AI is the translator.
Shirley said the goal was to show “how advanced AI technologies can unify fleet maintenance operations across a region as complex and fragmented as APAC.”
The starting point, according to Shirley, is that vehicle maintenance data is more universal than many people assume. Every vehicle has a registration plate, and in many countries that can be linked to a VIN. From there, the vehicle can be connected to a model identifier, which unlocks manufacturer service and repair data.
“The model ID is the key to the service and repair data from every manufacturer,” Shirley said.
That matters because once a fleet system knows the vehicle, the required job and the expected labour time, it can begin to calculate whether a quote is reasonable.
“The secret here is if you know what needs to be done, and you know how long it takes, then you can ask a repairer, ‘What is your agreed fleet rate, and what are the parts cost, and are there any agreed markups?’” Shirley said. “And you can calculate the cost of any service and repair in real-time without a human.”
For Fleet Managers and Fleet Coordinators, this is where the keynote moved from theory to practical operations. Shirley explained that a maintenance event is not a single action. It can include asset identification, vehicle configuration, diagnosis, scheduled servicing, unplanned repair, tyre replacement, procedure lookup, parts pricing, quote generation, approval, warranty handling, invoicing and compliance reporting.
The problem is that much of this work still arrives as unstructured data.
“Unstructured data is human voice. Yeah, a telephone call, but it could also be a fax, an invoice, an email, a text message,” Shirley said.
That is difficult for traditional systems because every service provider may use different software, and every market may have different processes, languages and local practices. AI changes the equation because it can interpret human context and convert it into a structured format that can be reviewed, measured and acted on.
However, Shirley also gave a clear warning about relying on large language models without controls.
“If you ask an LLM, whichever one’s your favourite, the same question three different ways, you’ll get three slightly different answers,” he said. “And that’s not practical when you’re talking about operational situations as critical as fleet maintenance and compliance.”
For Shirley, the solution is not just AI, but AI connected to controlled data and repeatable business rules. He described this as RAG-enabled AI, where unstructured inputs can be interpreted and then mapped back to a reliable data source.
During the keynote, Shirley introduced “MIC”, Fleet Guru’s AI maintenance co-pilot. In the live demonstration, Mick took a maintenance authorisation request by voice, confirmed the vehicle registration, identified the vehicle, asked for missing details, and submitted a structured maintenance authorisation through Fleet Guru.
The demonstration then moved across languages. Kota Tsuda, General Manager at Fleetguru Japan, began a service request in Japanese, before the authorisation continued in French. MIC carried the context across both languages and completed the request.
“What you just saw then was what we call a service agent, which can basically interpret unstructured data via voice, text, fax, or an email and convert it into structured data,” Shirley said.
For APAC fleets, the multilingual capability is the important part. The region includes mature and developing fleet markets, different repair networks, different administrative systems and many languages. Shirley said the technology being demonstrated could cover “ninety-two percent of Asia”, including major languages and Indian dialects.
The potential benefit is not only smoother communication. It is also cost control.
In a second demonstration, Shirley asked MIC to act as if they was a Fleet Manager for a fleet of 850 vehicles. MIC identified 12 maintenance items that had been flagged as outliers. One example was a brake pad part on a 2021 Iveco Daily, where the quoted price was higher than the peer median across similar jobs.
The system did not automatically reject the quote. Instead, it identified the variance and suggested a clarifying call before approval. It also highlighted a labour time concern on a 2018 Hino 300 Series, and a refrigeration unit repair quote that was around 30% above the peer median.
This is where AI could support Fleet Managers without removing their judgement. Routine jobs that meet business rules can be approved automatically, while exceptions can be sent to people for review.
“The human is the trust layer,” Shirley said during the post-keynote discussion. “You can’t sue the code.”
This is an important distinction for fleet buyers and Procurement Managers assessing AI claims. The value is not in replacing experienced people. The value is in scaling their capacity, reducing repetitive administration, and giving them better evidence before decisions are made.
“It’s not whether or not the human will stay in the loop. They absolutely will,” Shirley said. “It’s whether or not a single person can manage 1,000 vehicles or 100,000 vehicles.”
Shirley said FleetGuru is already seeing average auto approvals of 35%, with some customers reaching 70%. The reason, he said, is that the system knows what the job should cost, whether the service is due, whether the scope is predictable, and whether the price is at or below the peer median.
For global and regional fleet management companies, the bigger opportunity may be cross-border consistency. Shirley said the same problems appear in different markets, whether in APAC or Europe. Fleets and fleet management companies may know how to operate locally, but multinational customers increasingly want visibility across regions.
“The cross-border visibility and reporting will be hugely attractive to multinational and global customers,” Shirley said.
He also argued that richer maintenance data could improve total cost of ownership analysis and sharpen lease pricing because the inputs become more accurate and consistent.
For Fleet Managers, one of the more practical takeaways from the keynote came during the interview after the presentation. When asked about predictive maintenance insights that can deliver immediate savings, Shirley pointed to a basic but often missed opportunity: combining workshop tasks to reduce downtime.
“Before a vehicle goes in for a service, check the rego, check the tyres, and make sure that we can’t do an inspection or tyres or some other procedural thing that would take the vehicle off the road at the same time as the scheduled maintenance,” he said.
He said downtime remains one of the largest hidden costs in fleet maintenance.
“Time off the road is the number one cost for vehicle downtime, so let’s just reduce the number of visits to workshops,” Shirley said.
The message for the fleet sector is that AI does not need to be abstract. It can answer very practical questions: Is this vehicle due for the work? Is the quote reasonable? Is the labour time outside the normal range? Can another task be completed while the vehicle is already off the road? Can a maintenance request in one language be converted into structured data that a regional hub can manage?
For APAC, where scale, language and market fragmentation all add complexity, the technology could become a way to standardise maintenance decision-making without forcing every market into the same operating model.
Shirley’s keynote also served as a reminder that AI is only as useful as the data and controls behind it. In fleet maintenance, the best outcomes are likely to come from systems that combine human judgement, structured vehicle data, repair history, business rules and multilingual interfaces.
AI may be the translator, but fleet knowledge still matters.
