From training judgment to fostering creativity: How AI is helping staff from these SMEs grow on the job

During a tax advisory assignment at accounting firm Accountingbpo, an AI tool confidently presented an analysis of tax agreements involving Singapore, Hong Kong and China.

But an accountant spotted a big mistake: The tool had cited provisions that did not exist in the relevant agreements. Even when challenged, it insisted its interpretation was correct.

Without the accountant’s professional expertise, the error might have gone unnoticed and resulted in incorrect advice to a client. The incident highlights an important reality of AI adoption: Even as machines take on more complex tasks, people remain essential to ensuring things are done right.

“An accountant who lacks sufficient technical knowledge might accept a convincing AI-generated answer, particularly when it appears to be supported by legal references,” says Accountingbpo founder and partner Hu Tianying.

Founded in Singapore in 1997, the firm operates in Singapore and Vietnam, with a representative office in Taiwan and a recently established Korea desk. Its 16 professionals serve around 400 clients, primarily small and medium-sized enterprises (SMEs).

Since June 2025, the firm has been developing its own AI applications to process financial documents, coordinate assignments and generate reports. It won the SME Innovation Excellence Award for its proprietary AI solution at the inaugural SME AI Impact Awards 2026.

Organised by the Infocomm Media Development Authority (IMDA) and Singapore Business Federation, the awards named 30 winners from 290 nominations across 21 industries.

Says IMDA chief executive Ng Cher Pong, who was on the judging panel: “What sets the winners apart is not the tool that they have adopted or how technically sophisticated the technology is, but how well AI is embedded into their business workflows and the positive difference it has made to their operations, employees and customers.”

While AI can tackle routine work, the Accountingbpo management recognises that younger or less experienced employees still need to develop the expertise to challenge its output.

To build that knowledge, junior accountants use AI to research technical questions within the Inland Revenue Authority of Singapore’s IRAS e-Tax Guide and the firm’s own resources, which lets them check its answers against authoritative sources.

In-house GST learning materials and quizzes sharpen their knowledge further. Senior colleagues also mentor them, and junior employees have more opportunities to interact with clients and understand how their businesses operate, rather than simply process transactions.

This not only helps them learn how transactions should be recorded but also why – the knowledge that enables them to question AI-generated answers.

GetGo Technologies has also found that AI works best with people guiding its use.

Founded in 2021, GetGo is Singapore’s largest carsharing business that receives a large volume of bookings daily for its fleet of over 3,000 vehicles. To ensure that the fleet is well maintained, GetGo requires drivers to take photos of the car’s exterior before and after each booking.

“This generates about 56,000 photos a day, making it manually intensive for our team to sift through,” says Ang Deyao, chief operating officer of GetGo.

“Tracking a single instance of damage could take days as the team would need to look through backdated photo records to find the booking where the damage first appeared.

The GetGo data, product and engineering teams started building an in-house solution in 2024 and deployed Project Desmond in May last year. The AI solution – which also earned GetGo the SME Innovation Excellence Award – combines technologies such as computer vision and neural networks, and was trained on a collection of GetGo vehicle photos.

Project Desmond reviews every submitted image, identifies potential damage and compares suspicious areas against photos from previous trips to determine whether the damage appears new.

Only potential new damage is shortlisted for GetGo’s review team, which makes the final call using human judgment. Confirmed cases are then passed to the fleet team for follow-up with the user and repairs.

Like a new employee, Project Desmond receives regular guidance from its human seniors to improve its accuracy.

“When Desmond mistakes raindrops or a reflection for damage, our reviewer marks the vehicle as clean,” says Ang. “When it misses real damage reported by our users or our operators, we flag that too. Our fleet team then labels these photos and we retrain Desmond on them.”

Deepening personal connection and creativity

AI can also give staff more time for work that relies on human understanding and creativity, as English tuition provider Lil’ but Mighty Education Centre has shown.

The centre – which has 40 teachers and staff serving more than 1,500 students across nine centres – AI has been incorporated into customer service, teaching materials and administrative work since 2023.

Its WhatsApp AI assistant handles routine enquiries from parents about programmes and class schedules, while other AI tools help staff draft replies. Together, these have cut parents’ waiting time from around two hours to 20 minutes. An AI-supported communication tracker also helps teachers keep track of conversations with parents.

For these efforts, it won the SME Adoption Excellence Award, which recognises firms deploying and customising off-the-shelf solutions.

But the difference was also felt by students. A mother had told a teacher that her son felt disheartened after finding an oral examination difficult.

Before his next lesson, the teacher used an AI-supported communication tracker to revisit the earlier conversation and plan how to encourage him. When the teacher checked in, the boy was surprised that she remembered how he had felt.

“That interaction mattered because it helped a new student feel understood and supported,” says Abellon.

Another teacher used the tracker with a Primary 6 student in his Primary School Leaving Examination year. His parent said he had become reluctant to finish his work, so rather than push for more revision, the teacher revisited her earlier conversations with the parent to understand what was going on.

When she asked how he felt, he opened up about worries his parent had not known about. She passed these on, so they could support him together.

“It was not AI that understood the child or helped him open up; it was the teacher,” says Abellon.

“AI simply made it easier for her to access the information she needed.”

That understanding of students also guides how educators use AI to develop teaching materials. The technology helps them bring their ideas to life, while they decide what will help children learn.

For example, teachers describe scenarios for oral and situational writing exercises and use generative AI to create accompanying images – work that once involved hours searching for photographs, assembling illustrations or working with graphic designers. They still decide what makes an engaging, relevant exercise and whether the visuals suit their students.

“AI is there to enhance what our educators create, not replace their creativity, expertise and understanding of how children learn,” says Abellon.

Learn more about the 30 winners of the inaugural SME AI Impact Awards here.