By KYN AI Advisory Team — AI implementation specialists, Singapore
IMDA's Singapore Digital Economy Report 2025 puts AI adoption among Singapore SMEs at 14.5% in 2024, up from 4.2% in 2023. Among larger firms the figure was 62.5%. The report says the SME rise came mainly from off-the-shelf generative AI tools. For an SME owner the reading is plain: most peers have not adopted AI, those that have mostly use general chat tools, and the gap is in connecting AI to company data and workflows, not in access to a tool.
What does the IMDA report say?
The figures below come from IMDA's Singapore Digital Economy Report 2025, which covers 2024 and compares with 2023.
| Group | AI adoption 2023 | AI adoption 2024 | |---|---|---| | SMEs | 4.2% | 14.5% | | Non-SMEs | 44.0% | 62.5% |
The report states that SME adoption "tripled", and that the increase was driven mainly by micro and small SMEs, most of which adopted off-the-shelf generative AI tools such as ChatGPT, Gemini, Midjourney and Copilot. Non-SME adoption rose by around 20 percentage points.
The report also defines its terms. AI is a set of tools that enable computers or machines to perform tasks that typically require human intelligence, with applications such as content generation, customer engagement, analytics and process automation. The adoption rate is the percentage of firms using AI under that definition, and it includes AI built into other software.
Later government speeches cite different series, for example an enterprise-wide figure for 2025. Do not mix series. This article uses only the 2024 figures in the report we opened.
What counts as "adopting AI"?
The definition is broad. A firm that uses a chat assistant to draft emails is an adopter. So is a firm that runs a custom system reading its invoices. The statistic does not separate them.
That matters for how you read your position. Being among the 14.5% says little about depth. Two questions give you more:
- Which tasks does AI do without a person retyping its output?
- Does the AI read your own records, such as your CRM, ledger or stock list, or only text a person pastes in?
If the answers are "none" and "pasted text only", you use AI as a writing aid. That has value, and it is also what most of the 14.5% do.
Why is the SME figure so far behind larger firms?
IMDA does not list causes in the figures we cite, so what follows is our view, labelled as such.
- Scale of the gain. A firm with 400 invoices a day gains more from automation than a firm with 40, so larger firms move first.
- Fixed cost of integration. Connecting an AI to a ledger, a CRM or a warehouse list takes effort whatever the firm's size. For a small firm that effort looks large against the gain.
- Data in the wrong places. SME data often lives in spreadsheets, inboxes and WhatsApp threads. An agent cannot act on what it cannot read.
- No owner. Larger firms have a person whose job includes this. In a 15-person firm the founder does, between other jobs.
None of this is a barrier to start. It shapes where to start.
What does the data suggest an SME should do first?
Our recommendation: begin with one repeated, rule-based task where an error is cheap to catch, and where the data already exists in one place.
Good first tasks:
- sorting inbound enquiries by type and routing them to a person;
- reading supplier bills and posting drafts for review;
- compiling a daily or weekly report from a system you already use;
- chasing overdue invoices at fixed intervals.
Poor first tasks:
- anything that sends money or files a return without a person checking;
- anything where the source data is missing or stale;
- anything that needs judgement about a relationship.
Our guide to where a Singapore SME should start with AI adoption covers the selection in more depth.
What does "off-the-shelf" miss?
Off-the-shelf generative AI tools are general. They know nothing of your price list, your customers or your approval rules, unless a person pastes them in each time. They also run under whatever login the user holds, and what goes into them leaves your control to a degree set by the vendor's terms.
A custom system differs in four ways:
| Point | Off-the-shelf chat tool | Custom agent system | |---|---|---| | Reads your records | Only what is pasted | Directly, with set permissions | | Acts without retyping | No | Yes, within rules | | Audit trail | Chat history | Every action logged with the actor | | Approval gates | None | Set per action type |
The difference is not intelligence. It is access, rules and records. For the security side see security risks before using AI for business.
How should you use the statistic in a decision?
Do not use 14.5% as a reason to hurry, or to wait. Use it as a check on assumptions. If a vendor says "everyone is adopting AI", the SME figure says otherwise. If a team member says "we are behind", the figure says most peers are at the same stage, and the useful comparison is against your own cost and error rates.
Use your own data instead:
- hours a week on repeated tasks, from a time log;
- errors found after the fact, from your corrections file;
- days from enquiry to first reply.
These three tell you where AI has a return in your company. An industry statistic cannot.
What would change the picture next year?
Three things are worth watching in the next IMDA report, and none can be predicted from the 2024 figures.
- Whether the SME rate keeps rising at the same pace. A tripling from a low base is easier than a second tripling.
- Whether the report separates depth of use. A count of firms using AI says little. A count of firms using AI on their own records would say more.
- Whether the gap to larger firms narrows. The 2024 gap is 48 percentage points (62.5% less 14.5%). If it narrows, the cause matters: did SMEs improve, or did larger firms level off?
Until then, treat the 2024 numbers as a snapshot of tool access, not of results. The report's own conclusion refers to adoption rates, not to outcomes, and an adoption rate does not say whether the AI paid back.
Where KYN fits
KYN builds the connected layer that general chat tools lack. For an insurance brokerage, we built outreach with one checkable first line at about S$0.02 per email, a reply desk at about S$0.01 per reply sorted, and a comparison from group benefits data to a client PDF at about S$0.03. A manufacturer cut eleven systems to three on one AWS account.
The first 14 days are a review of your stack and interviews with the team. A full build ships in under 90 days, in a cloud account in your name. Agents run under their own restricted accounts, and outbound messages, money, filings and deletions wait for a named person. See infrastructure and security.
Frequently asked questions
What percentage of Singapore SMEs use AI?
IMDA's Singapore Digital Economy Report 2025 gives 14.5% for 2024, up from 4.2% in 2023. Check IMDA for a newer edition before you cite the figure.
Is the IMDA figure about custom AI or chat tools?
It covers both. The report says the SME increase was driven mainly by off-the-shelf generative AI tools, and its definition includes AI built into other software. The figure does not show how deeply firms use AI.
Should a small firm wait until more SMEs adopt AI?
Waiting for peers is not a plan. Decide by your own numbers: hours on repeated work, error rate and reply time. If those are low, wait. If they are high, start with one task.
Checked 30 October 2026.
Curious whether Marketing agents fits your business? Talk to KYN on WhatsApp — no forms, just a conversation.
