If 10,000 Workers Are Reskilled, Have We Succeeded?

As AI reshapes work, perhaps the harder question is not how many people we train, but how we know whether that training genuinely changed their future.

One part of this year's National Day Rally stayed with me: as Singapore embraces AI and other new technologies, we also need to help workers adapt as jobs change. As someone who works in both analytics and training, it made me think about a deceptively simple question.

How will we know whether our reskilling efforts actually worked?

Imagine that a major workforce programme publishes the following results:

10,000 workers enrolled. 9,200 completed their training. 8,500 received certification.

Those would sound like impressive numbers, and rightly so. They tell us that people participated, most completed the programme and a large proportion achieved the required standard.

But what exactly have we measured?

We know how many people were trained and how many completed the programme. We may even know how satisfied they were with the training. What we do not yet know is what happened to them afterwards.

Did the training help them move into meaningful employment? Were their new skills actually used? Did the transition give them a sustainable career path and the opportunity to rebuild or improve their earning power over time?

These are harder questions, but they are closer to the outcome we actually care about.

This illustrates one of the most common problems in analytics: we measure what is easy to count and can gradually begin treating it as though it were the thing we set out to achieve.

 Training numbers tell us what we delivered. They do not necessarily tell us what changed.

Training Is an Output. Employability Is an Outcome.

There is nothing wrong with measuring enrolments, completions and certifications. Programme managers need these numbers because they tell us whether resources are being used, whether participants complete what they started and whether programmes are operating as intended.

But these are primarily measures of activity and output.

For workforce reskilling, the outcome we ultimately care about is something more substantial: whether people can make a successful and sustainable transition into meaningful work in a changing economy.

Consider two mid-career professionals who complete exactly the same data analytics programme. Both attend every session, pass the assessment and receive the same certificate. From the perspective of the training dashboard, both are successful completions.

Their experiences afterwards, however, may be very different.

One might move into a new role where the skills are being used, build on previous experience and begin developing a new career path. The other might struggle to find an appropriate opportunity and eventually accept a role that makes little use of either their previous experience or newly acquired capabilities.

The dashboard records two successful completions, but their career outcomes tell us two very different stories. That is the distinction between measuring whether a programme was delivered successfully and measuring whether it achieved its purpose.

Getting a Job Is Only Part of the Story

Singapore's labour market remains relatively resilient. In the first quarter of 2026, total employment continued to grow, unemployment remained low and the six-month re-entry rate among retrenched residents improved from 57.4% to 60.7%. There were also more job vacancies than unemployed persons. (stats.mom.gov.sg)

These are encouraging and useful measures. They tell us something important about whether displaced workers are finding their way back into employment.

But measurement always needs to begin with the question we are trying to answer.

If the question is, "Did this person find another job?", re-entry into employment is a perfectly reasonable measure.

If the question is, "Did this person make a successful career transition?", it is only part of the answer.

Someone may return to employment quickly but move into a role with substantially less responsibility, limited use of their experience or few opportunities for future progression. Another person may take longer to find work but eventually move into a role that creates a sustainable new career.

A re-employment statistic is not designed to distinguish between those experiences, nor should we expect it to. No single measure can tell us everything. The problem begins when we allow one measure to stand in for a much larger outcome.

Perhaps We Should Measure the Journey

Instead of treating reskilling as a single event, it may be more useful to think about it as a journey:

Training → Transition → Employment → Sustainability → Progression

Training tells us whether people acquired relevant new capabilities. Transition considers whether those capabilities helped them move from their previous role or industry into another one, while employment looks at whether they secured work and how long that took.

The measurement should not stop there.

Sustainability asks whether the person remains successfully employed over time. Progression goes further by considering whether the transition has created a platform from which skills, responsibilities, earning power and career prospects can continue to grow.

Seen this way, course completion is not an unimportant metric. It is simply an early indicator in a much longer chain. A more complete view might therefore combine completion and certification with measures such as time to re-employment, relevance of the new role, utilisation of newly acquired skills, wage recovery, employment retention and subsequent career progression.

Not every program needs every measure. What matters is that the measures reflect what the program was created to achieve. We should define success first, then decide what to measure. Not choose what is easy to measure and allow that to define success.

Reskilling is not a single event. The more meaningful question is what happens to the worker after the training ends.

Singapore Is Already Moving Towards Outcomes

Importantly, this is not an argument that Singapore's workforce programs measure only course attendance or certification. In fact, recent data show that employment outcomes are already being tracked in more meaningful ways.

In August 2026, the Ministry of Manpower reported that about nine in ten participants in Career Conversion Programs remained employed for at least 24 months after the program, while about seven in ten earned more than their last drawn salaries. For SkillsFuture Career Transition Programs, about half of trainees found new roles or employment within six months of completing their course. (mom.gov.sg)

There is encouraging evidence for older workers too. Over the last five years, close to nine in ten Career Conversion Program participants aged 51 and above remained employed 24 months after embarking on the program. For the Mid-Career Pathways Program, more than six in ten participants aged 51 and above found employment within six months after completing or exiting the program. (mom.gov.sg)

These measures tell us considerably more than participation alone because they begin to answer the more important question: What happened to people after we trained them?

They also illustrate something important about good analytics. The answer is rarely to abandon an existing metric simply because it is incomplete. Completion rates still matter, as do participation and certification. What we need is a set of measures that collectively gives us a better picture of the outcome we care about.

The Metric Is Not the Goal

This measurement challenge goes far beyond workforce policy.

We encounter the same issue inside organisations all the time. A learning department measures employees trained. A sales team measures leads generated. A customer service team measures response time. A hospital measures waiting time. A digital transformation team measures adoption.

Every one of those metrics can be useful, but none necessarily represents the ultimate outcome.

An organisation does not invest in training because it wants more certificates. It does so because it wants people to become more capable. A sales team does not really want leads. It wants profitable customers. And a digital transformation programme does not ultimately succeed because people logged into a new system. It succeeds when the technology helps people or the organisation perform better.

This is closely related to a familiar principle in measurement: when a measure becomes a target, people naturally begin optimising the measure, sometimes at the expense of the purpose it was meant to represent.

The problem is not the KPI. The problem is forgetting why we created it.

A useful KPI becomes dangerous only when we expect it to tell us more than it was designed to tell us.

This Is Where Analytics Should Help

Good analytics should do more than tell us whether a KPI went up or down. It should help us determine whether the KPI still represents the outcome we care about.

If thousands of people complete training, that achievement deserves recognition. But analytics should also help us follow what happens afterwards by examining the types of roles people enter, how their earning power develops, whether their new skills are being used and which groups make successful transitions compared with those who continue to struggle.

Answering those questions may require longitudinal data rather than a dashboard snapshot. It may mean connecting information across programmes, employers and different periods of a person's career. Some questions may even require qualitative evidence because not every meaningful career outcome can be reduced neatly to a number.

But difficulty of measurement should not determine importance.

Sometimes the things that matter most are precisely the things that are hardest to count.

This becomes even more important as AI changes the nature of work. MOM's latest labour-market data suggest that AI is currently affecting how jobs are performed more than whether jobs disappear altogether. Firms adopting AI were more likely to report redesigning job functions than reducing headcount or hiring because of AI. (stats.mom.gov.sg)

Workforce transformation may therefore not always look like someone losing one job and training for an entirely different one. Increasingly, existing jobs may change around us, requiring people to acquire new capabilities while adapting and building on the experience they already have.

That makes successful reskilling harder to measure, but also more important to understand.

Define Success Before We Measure It

As AI changes the nature of work, Singapore will need to continue investing in training and reskilling. The National Day Rally's emphasis on helping workers adapt while embracing technological change is therefore timely. (pmo.gov.sg)

But the number of people trained should never become the destination.

What matters is what happens afterwards. Did people make successful transitions into meaningful work? Are their new skills being used? Have they retained or rebuilt their earning power? Are they developing careers that can continue adapting as industries and technologies change?

These questions are harder to answer than counting course completions because they require us to follow outcomes over time and recognise that a successful career transition has more than one dimension.

But that is precisely the point.

Good analytics is not about measuring everything we can. It is about measuring what helps us understand whether we achieved what we set out to achieve.

So perhaps the question is not simply how many Singaporeans we can reskill for the age of AI.

It is how we will know when reskilling has genuinely changed someone's future.