UOB survey finds 65% AI adoption but only 15% advanced capability

What the UOB numbers show

A survey of business owners across the region by United Overseas Bank (UOB), reported in an analysis published on dev.to, found that 65% of businesses have adopted AI in some form, while only 15% have reached what the study classifies as advanced capability. On those figures, roughly four in five companies using AI sit somewhere between an initial experiment and a genuine change in how the business runs.

The barriers respondents named are telling. According to the dev.to write-up of the study, 47% cited data and system readiness, the same share that cited funding, with 39% pointing to talent. The author's argument is that funding and talent get the attention, but data readiness is the constraint that decides whether a pilot ever becomes an operational process.

Why pilots succeed and production fails

The core of the analysis is a pattern familiar to anyone who has run a proof of concept. A small dataset — say, a few months of sales records — is tidied up by the one person who understands its quirks, loaded into a chatbot, and returns an impressive answer in seconds. Extending that same exercise to every product line on a weekly schedule falls apart, not because the model got worse, but because nobody can repeat the manual data cleanup at scale. A pilot tests AI against a curated sample; routine use requires an underlying system.

The UOB data bears this out in structural terms. The study found a 28-point gap in digitalisation success rates between large and small enterprises, and in Singapore only 69% of small enterprises are digitalised against 93% of large ones. The dev.to piece reads this not as large companies having better AI, but as their already having systems where transactional data resides.

What data readiness actually involves

Stripped of jargon, the article defines readiness as four unglamorous conditions: one record per real-world entity, so a customer is not three near-duplicate lines no tool can aggregate; figures that update without someone retyping a spreadsheet; commercial knowledge that survives staff turnover rather than living in one person's head; and systems that can export or connect programmatically instead of relying on copy-paste into a chat window.

The sequence the author recommends follows from that: fix master data first, consolidate transactions so that a sale, a stock movement and an invoice register as one event — the role an ERP system plays — and only then place AI on top, where it can forecast from real history, flag margin drift and draft routine purchase orders. One caveat worth noting: the piece is written by an ERP consultant and closes by promoting an ERP readiness tool, so its framing of ERP as the foundation carries a commercial interest. The survey figures themselves are UOB's.

Reported gains and business priorities

The study is not all gap. According to the same data, 47% of businesses reported cost reductions from AI and 46% reported productivity gains. The analysis attributes those results to companies that already had infrastructure to connect the technology to — the tools were equally available, but the underlying data foundations were not. Elsewhere in the survey, one in three businesses named rising costs as their top challenge, and improving profitability narrowly outranked growing sales as the leading priority.

Why it matters

The UOB numbers quantify something most businesses have only felt anecdotally: adoption is the easy half, capability is the hard half. For small firms weighing AI budgets, the data suggests the binding constraint is often not the model or even hiring, but whether records are consistent, current and extractable. The 69%-versus-93% digitalisation split implies the AI capability divide may largely replay an existing infrastructure divide — and that companies who fix their data foundations first may find that adding AI becomes a small, incremental step rather than a stretch they repeatedly fall short of.