Singapore's SME AI adoption rate was 14.5% in 2024, up from 4.2% in 2023, according to IMDA's Singapore Digital Economy Report 2025. That is the number most articles quote. It is also incomplete.
A separate DBS Business Pulse Check survey of 730 Singapore SMEs, fieldwork run December 2025 to January 2026, found 67% of respondents already applying some form of AI in their work. Only 12% called it fully integrated across operations. Both numbers are accurate. They measure different things: whether the business has adopted AI as a structured practice, and whether anyone inside the business is using AI tools at all.
The two figures side by side
- IMDA14.5% of SMEs, 2024. Tripled from 4.2% in 2023. Measures structured, business-level AI adoption.
- DBS67% of 730 SMEs surveyed, Dec 2025–Jan 2026. Measures whether anyone in the business uses AI at all, including informal use.
- GapOnly 12% of the DBS sample called their AI use fully integrated. Most of the 67% is individual staff experimenting, not a business process.
Two numbers, same country
If you searched for Singapore's SME AI adoption rate at any point in the last year, you likely landed on one of two figures, and they do not agree. One says adoption is still low. The other says most SMEs are already using AI. Read in isolation, either headline supports a different argument about how ready Singapore SMEs are.
Read together, they describe something more specific and more useful: a country where individual AI use has spread quickly, but business-level ownership of that use has not caught up.
What the 14.5% measures
IMDA's Singapore Digital Economy Report 2025, released in October 2025, tracks AI adoption as part of its broader measure of digital solution adoption across sectors. On this measure, SME AI adoption tripled in a single year: from 4.2% in 2023 to 14.5% in 2024. Non-SMEs moved further over the same period, from 44% to 62.5%.
A tripling is a genuinely fast rate of change. It also means that, as of the most recent data, roughly 85 in every 100 Singapore SMEs had not adopted AI on this definition. This is the figure that gets cited when the story is “SMEs are behind.”
What the 67% measures
DBS surveyed 730 companies between December 2025 and January 2026 for its Business Pulse Check. About 67% of respondents said they were already applying some form of AI in their work processes. Adoption was uneven by sector: information and communications and electronics manufacturing firms led, wholesale and trade firms lagged.
The detail that matters most sits inside that 67%: only 12% described their AI use as fully integrated across operations. The rest is somewhere between one team experimenting and a founder using AI for a single recurring task. It is real use. It is not yet a system the business runs on.
A business can score zero on formal adoption and still have several employees using AI every day.
Why the two numbers diverge
Individual AI use has almost no barrier to entry. A staff member opens a free or low-cost tool, gets useful output on a task, and keeps using it. No budget approval, no training, no one else in the business needs to know. That is enough to show up in a survey question asking whether anyone is using AI.
Structured adoption is a different kind of decision. Someone has to choose which tools the business standardises on, set basic rules for what information can go into a prompt, and make sure the people using the output know how to check it. None of that happens automatically once a few employees start using ChatGPT on their own.
In conversations with SME marketing teams, this is the pattern that shows up most often: someone on the team already uses AI for a task or two, and nobody has decided whether that is the business's approach or one person's habit. The two numbers are describing that exact moment, at national scale. It also explains why the gains are so uneven: as we found looking at AI productivity inside marketing teams, the people who benefit most are the ones who already had enough domain knowledge to judge what the tool gave them, not just access to the tool itself.
What SMEs say would help
In the same DBS survey, respondents pointed to three forms of support as most useful for moving past informal use: financial support and grants, expert guidance on how to apply AI effectively, and partnerships with technology providers who understand their sector.
None of the three is a technology problem. A grant does not tell a business which tool to standardise on. A partnership does not decide what data policy to set. What SMEs are asking for, in different words, is someone to help them turn individual tool use into a decision the business has actually made.
What to do with this
If you run an SME marketing team, the useful question is not “have we adopted AI.” It is closer to the DBS question: who on the team is already using AI, for what, and does anyone know.
Start by finding out. Ask each person on the team what they already use AI for day to day. You will likely find more activity than you expected, scattered across two or three tools with no shared standard for how the output gets checked.
From there, the decision is not whether to adopt AI. It already has been, informally, by whoever on the team started using it first. The decision is whether the business owns that choice or keeps leaving it to whoever happens to try a tool next.
Common questions
Is Singapore's SME AI adoption rate really only 14.5%?
That figure comes from IMDA's Singapore Digital Economy Report 2025, and it measures formal AI adoption: businesses that have built AI into a defined process, tool, or workflow. It tripled from 4.2% in 2023 to 14.5% in 2024. It is accurate, but it is not the full picture of how much AI activity is happening inside SMEs.
Why do the IMDA and DBS numbers look so different?
They are measuring different behaviour. IMDA's 14.5% tracks structured, business-level AI adoption. DBS's 67% (from a survey of 730 SMEs, fieldwork December 2025 to January 2026) tracks whether anyone in the business is using AI tools at all, including informal use by individual staff. A business can score zero on the first measure and still have several employees using ChatGPT daily.
Does having staff use ChatGPT count as AI adoption?
It counts as AI use. It does not count as adoption in the sense that matters for consistency, data handling, or return on the time spent. If nobody in the business owns the decision of which tools are used, how outputs are checked, and what data goes into a prompt, the business has AI activity but not an AI process.
What's actually stopping SMEs from moving from informal use to structured adoption?
Individual use is low-cost: a free or cheap tool, no approval needed, no training required. Structured adoption requires someone to decide which tools the business standardises on, set basic rules for what data can go into a prompt, and train the team to use the output well. That is a resourcing and ownership decision, not a technology one, which is why it lags behind informal use.
Where can Singapore SMEs get help adopting AI properly?
DBS's own survey respondents pointed to financial support, expert guidance on applying AI effectively, and partnerships with technology providers as the most useful forms of help. Hackalogy's advisory sessions and training programme are built around the same gap: helping a team move from ad hoc tool use to a process it can actually rely on.
AI Marketing Advisory
Turn informal AI use into a process your team can rely on.
One session to map out which tools your team should standardise on, what data rules to set, and how to check the output. No retainer required.
Sources
- 1. Infocomm Media Development Authority, “Singapore Digital Economy Report 2025,” October 2025. imda.gov.sg
- 2. CRN Asia, “Singapore's digital economy surges to 18.6% of GDP as AI adoption triples among SMEs,” October 2025. crnasia.com
- 3. DBS Bank, “DBS Survey: 8 in 10 SMEs prioritise overseas expansion in 2026 to drive growth amid market volatility,” 2026. dbs.com
Observations on SME marketing teams are drawn from Hackalogy and Neo360 client conversations, 2025–2026, and reflect patterns seen across multiple teams, not a single case study.