Singapore banks bet on AI while racing to reskill their workforce

Singapore banks are all-in on AI. HR has to make sure people come with it

AI is moving from pilot to production across Singapore's banks. Call centres, KYC, transaction monitoring, analytics, and even software delivery are changing fast.

The talent question is now the business question: how do you lift productivity without losing your people? A January forecast suggested global banks could cut up to 200,000 jobs in three to five years. DBS, for instance, plans to trim about 4,000 contract and temp roles across 19 markets over three years as AI scales - while creating 1,000 AI-driven roles.

Where automation hits first

• Repetitive, rules-based work: data entry, document checks, and some teller and customer service tasks.
• KYC and transaction monitoring are increasingly machine-supported, lifting the bar for many back-office roles.
• Parts of junior analyst work (data crunching, first-pass analysis) are now AI-assisted.
• Temp and contract roles that don't need specialist skills will likely decline as automation smooths seasonal peaks.

This will shrink classic entry-level pipelines. As Gaurav Kwatra of Oliver Wyman notes, banks will still protect their future leadership bench - but the shape of early careers changes.

Augment, don't replace

"With all technologies, there's always talk that it is going to eliminate lots of roles, but history has proven us all wrong," says Yap Aye Wee, OCBC's head of learning and transformation for group HR. Roles disappear; new ones show up.

UOB has kept headcount steady the last two years while doing more with technology, says Dean Tong, group head of HR. Standard Chartered's Wong Yang-Sheng is clear: Gen AI boosts productivity, but human judgement still carries the decision.

Real examples:

• OCBC contact centre: speech-to-text cuts manual transcription from hours to minutes, freeing time for call analysis. Human review still matters for tone and sentiment.
• UOB call centres: Gen AI supports agents now; simpler calls will go to machines, complex calls to people.
• Internal co-pilots: OCBC GPT, DBS-GPT, SC GPT, Citi AI help staff summarise, search, and draft with oversight.
• Citi developers: AI automates hundreds of thousands of code reviews weekly, saving an estimated 100,000 hours a week that are reinvested into innovation.

What HR needs to do next

• Map roles to tasks, not titles. Break jobs into task inventories. Mark tasks as automate, co-pilot, or human-led. Redesign roles around the new mix.
• Refresh job architecture. Add AI literacy, data fluency, prompt skills, and model oversight into competencies and career paths.
• Rebuild the entry-level pipeline. Expect fewer low-skill openings. Create apprenticeships, rotations, and projects where juniors learn judgement with AI at their side and a human mentor.
• Upskill at scale. Provide bank-wide AI foundations (including deepfake awareness). Deliver role-specific training just in time. Train leaders to both perform and drive change.
• Build internal academies. Follow UOB's lead: multi-year programmes with certifications, placements, and pathways for non-tech talent into tech roles.
• Hire for trust and thinking. Let candidates use AI in take-home tasks. Assess how they explain choices, verify outputs, and own the result. Values screening matters.
• Design human-in-the-loop. Define escalation rules, quality gates, and accountability for co-pilot outputs across frontline, risk, and ops.
• Stand up AI governance. Establish model risk controls, audit trails, privacy and data minimisation, bias testing, and content provenance. Align to MAS FEAT principles.
• Measure and reinvest savings. Track metrics like call handle time (DBS cut up to 20%), first-contact resolution, cycle time, defect rates. Reinvest saved hours in customer work and skill growth.

Skills banks are prioritising

• Technical: data literacy, prompt skills, analytics, model monitoring, automation tooling, secure AI usage.
• Human: critical thinking, ethical judgement, customer empathy, change resilience, communication.
• Leadership: set clear use cases, remove blockers, coach teams to use AI well, and uphold risk standards.

As OCBC's Yap advises new entrants: treat AI as a tool, not a master. Pair it with a human mentor to learn nuance that machines miss.

Operating model: five workforces to plan for

Accenture's Nesan Govender breaks the future bank into five interdependent workforces: human-led roles, human-machine collaboration, robotic automation, Gen AI, and fully autonomous agentic AI. HR should define how each is supervised, measured, and rewarded - and how people move across them.

Policy notes for Singapore HR teams

• Be a fast follower, safely. As UOB's Tong notes, first movers take outsized risk. Move quickly where controls are clear; stage bigger rollouts.
• Privacy and data residency. Clarify which data can be used for training vs. inference. Prefer private, logged co-pilots over open tools.
• Regulator alignment. Document explainability, fairness, accountability, and transparency practices. Build audit-ready evidence from day one.

90-day HR action plan

• Days 0-30: Pick three high-volume use cases (e.g., call handling, KYC checks, MI reporting). Create a skills matrix and risk checklist. Set metrics and guardrails.
• Days 31-60: Launch role-based training for impacted teams. Pilot human-in-the-loop workflows. Update job descriptions and internal mobility paths.
• Days 61-90: Scale what works. Publish productivity and quality gains. Reassign freed capacity to higher-value work and advanced training.

Why this matters for HR

The prize is meaningful: Deloitte estimates banks that execute well can improve cost-to-income ratios by 5-15% over five years. The risk is clear: hollowed entry levels and skill gaps if you delay.

The banks that win treat AI as an enhancer and invest in people who can use it well - with judgement, ethics, and accountability.

Helpful resources

• MAS FEAT principles for responsible AI
• Complete AI Training: courses by job