Independent commentary based on publicly available sources. MYAICO was not involved in this project and is not affiliated with or endorsed by Grab, OpenAI or Anthropic.
For many SME owners, the challenge is finding time to act on the information already available. Customer enquiries, daily operations and marketing decisions compete for attention.
AI becomes useful when it helps with a specific part of that work. Grab’s AI Merchant Assistant offers a relevant example of this approach in Southeast Asia.
The industry example
In April 2025, Grab announced an AI Merchant Assistant integrated into the GrabMerchant app. Developed in collaboration with OpenAI and Anthropic, it was designed to provide merchants with personalised business guidance and help with everyday tasks, including menu updates and advertising campaigns.
Grab described the assistant as a way to make business support more accessible to small merchants who may have limited time and resources for analysing their operations. Read Grab’s announcement.
For MYAICO, the useful lesson is the connection between AI assistance and a clear business need. An SME can explore that principle through a much smaller, focused project.
What could this look like in your business?
The following are MYAICO’s illustrative applications. They are not descriptions of Grab’s implementation or completed MYAICO customer projects.
For a retailer: preparing customer replies
An assistant could draft answers using an approved product catalogue, delivery information and returns policy. Staff would check the response and handle exceptions, such as disputed refunds or unusual requests.
For a distributor: finding product information
An internal assistant could help employees locate specifications and relevant documents. Answers would link back to their sources, while staff confirm suitability, pricing and availability before advising a customer.
For a service business: organising incoming enquiries
AI could summarise a customer’s request, identify missing information and prepare a follow-up draft. The team would retain responsibility for recommendations, quotations and commitments.
Each example starts with a defined task and a clear role for the person using the assistant.
Start with one workflow
MYAICO recommends choosing a task that happens regularly and has a result you can evaluate.
“Help our team prepare answers to delivery enquiries” is a clearer starting point than “use AI to improve customer service.”
Before selecting a tool, identify:
- Who performs the task today.
- Which information they need.
- Where delays or repeated work occur.
- Which decisions require human approval.
- What a useful improvement would look like.
This gives the project a practical purpose and a basis for judging whether it works.
Get the information ready
An assistant needs reliable business information to produce useful answers.
For an initial pilot, that might mean a small, approved collection of product details, procedures or frequently asked questions. Assign someone to keep those materials current and decide who should have access.
The workflow should also make missing information visible. When the assistant cannot support an answer, it should direct the user to a person rather than invent a response.
Keep people responsible
Define what the assistant may do and what needs review.
Preparing a draft is different from sending a customer message. Suggesting a next step is different from approving a refund or committing to a price.
For an early pilot, MYAICO recommends keeping customer-facing messages and consequential actions subject to human approval. That creates an opportunity to learn where the assistant is useful and where it needs improvement.
Measure usefulness before expanding
Test the workflow against representative tasks, including difficult questions and incomplete information.
Useful measures could include:
- Time spent completing the task, including review and corrections.
- Accuracy against approved information.
- How often a person needs to take over.
- Feedback from the employees using the assistant.
These are proposed evaluation measures, not promised outcomes. Expand the workflow only when the results justify doing so.
MYAICO’s perspective
A practical AI project begins with understanding the work.
For SMEs in Singapore and ASEAN, a focused assistant may be worth exploring when there is a recurring task, reliable information and a clear way to assess its usefulness. The appropriate solution will depend on the business’s systems, resources and needs.
Practical AI. Meaningful Progress.
Source: Grab, “Grab deploys agentic AI to empower merchants and driver-partners,” 9 April 2025. The external example is summarised in original wording. The proposed SME applications and pilot recommendations are MYAICO’s analysis; no equivalent results are implied.