Singapore's AI Strategy Needs Evidence From Frontline Workflows, Says Behavioural Scientist

Singapore's refreshed national AI strategy has all the top-down ingredients a small country could ask for: ten priorities, a new National AI Council steering the agenda, and A*STAR's freshly launched Institute of Advanced Intelligence and Computing pulling together AI, data and high-performance computing research. But behavioural scientist Gleb Tsipursky, who spent hundreds of consulting projects studying how organisations actually adopt AI, argues the decisive work happens far from the strategy documents — at the frontline, inside recurring workflows.

Writing in GovInsider as a guest columnist, Tsipursky points to Gallup's May 2026 workplace data as evidence of the gap between activity and transformation: 65 per cent of employees in AI-enabled organisations say the technology has improved their productivity, yet only 14 per cent strongly agree it has transformed how work gets done across the organisation. Individual gains, he warns, do not automatically become institutional gains — and counting licences, prompts or training completions mistakes activity for progress.

His prescription for the Singapore public service is five concrete habits: start with work that happens often enough to measure (summarising case files, answering common citizen queries, checking forms); assign a named human owner to every AI-supported workflow; treat frontline employees as co-designers rather than recipients; run pilots in weeks with short feedback loops instead of year-long transformation programmes; and measure outcomes citizens can actually feel — cycle time, errors, rework, escalation rates.

None of this is abstract governance theory. The Public Service has committed to transformation and better citizen services, and agencies like the health clusters are already shipping production AI tools at speed. The risk Tsipursky identifies is that enthusiasm distorts evaluation: without a baseline, almost any pilot can be called a success because people remember the impressive output and forget the time spent correcting failures.

Why it matters for Singapore: The country's AI advantage has never been raw compute — it is the density of digital institutions and a public service that actually implements. This column lands at a moment when the National AI Council is deciding what "AI for the public good" means in practice. If agencies adopt the frontline-evidence discipline — baselines, ownership, co-design, fast feedback — the national strategy converts into services Singaporeans can feel. If they don't, the strategy risks being a set of priorities that never survive contact with a caseworker's desk.