AI Productivity at Work Is Real.
The People Who Benefit Know Their Job First.

Singapore's AI Council is right to push for AI adoption at work. Marketers who use generative AI for image creation compress a two-hour brief-to-output process into four minutes, and copywriters generate multiple body variants in the time it used to take to draft one. Those time savings are genuine.

But at Neo360, my marketing agency, I keep watching a pattern the productivity numbers do not capture. The people who benefit most from AI tools are the ones who already understand their domain well enough to evaluate what the tool gives them. People without that understanding are getting faster at producing work they cannot assess.

What the AI Council is pushing for

Singapore's National AI Council has been making the case for companies to adopt AI tools for productivity. A June 2026 commentary in Channel NewsAsia put the logic directly: businesses that integrate AI into production workflows gain a real time advantage over those that do not. The argument is correct.

There is genuine evidence that AI reduces the time cost of production in marketing. Drafting copy and generating image options now take minutes where they used to take hours. A team that adopts these tools will move faster.

Where the productivity frame falls short is in distinguishing faster from better. More output per hour and better output per hour are different goals, and which one you achieve depends on something the productivity numbers do not address.

The copywriting case

Take copywriting. A junior marketer using AI generates five ad copy variants in two minutes. The same work might have taken an hour before. That saving is real.

The difficulty starts with what comes next. Selecting the right variant and knowing when to discard all five requires an understanding of what makes copy perform. If you have not studied what a strong hook looks like, you will pick whichever variant sounds most professional. AI is trained to produce copy that sounds like copy. That is not the same thing as copy that converts.

At Neo360, I have reviewed AI-assisted copy from clients who adopted these tools after seeing the time savings. Most of it clears the smooth and professional bar immediately. Very little of it would hold up in a live campaign. The clients who produce effective AI-assisted copy are the ones who already had a clear sense of what good copy looks like before they picked up the tool.

If you cannot identify what a strong hook does differently from a weak one, you cannot tell whether the AI gave you a good one. That judgment comes from time with the craft: studying campaigns that performed and understanding what made them work.

The ad campaign case

Ad campaign management is where the gap becomes more expensive. AI tools now generate performance summaries that would have taken an analyst most of a day to prepare, and surface audience and bidding recommendations with supporting logic.

Each of those recommendations needs to be evaluated against knowledge the AI does not have: how this specific client acquired their best customers and what they have already tested.

An AI might recommend expanding to a lookalike audience built from your client's highest-value customers. The recommendation will look reasonable in the data. What the AI does not know is that this client's highest-value customers came through word-of-mouth referrals in a narrow professional segment, and every previous lookalike expansion converted at roughly a third of the original rate. The marketer who knows that acquisition history will question the recommendation. Without that background, the most natural move is to run the audience.

The split I keep seeing

Two marketers can use identical tools and arrive at very different places six months later.

One uses AI to move faster on production, then applies domain judgment to filter and refine the output. Their capability grows alongside the speed because they stay engaged with the quality judgment at every step.

The other uses AI to replace the judgment step. They accept outputs that sound plausible and follow recommendations without working through the underlying reasoning. Their output volume goes up. Their ability to assess whether it is any good erodes, because they exercise that judgment less often.

The difference is not about which tool they use or how frequently. It is whether domain knowledge is present to act as a filter on what the tool gives them.

This is also where I think the discussion around AI replacing jobs misses the point for most marketing roles. AI does not replace the person with domain knowledge. It reduces the value of the person without it. That distinction matters if you are thinking about where to spend your development time.

What to do with this

If you are building AI capability in a marketing team, the useful question first is not which tools to adopt. It is what the person using them already understands about the domain.

For copywriters: before using AI for finished output, ask whether you can identify what makes a specific piece of copy effective. If you cannot rewrite a weak headline without prompting, use AI for ideation and research rather than output, and build that judgment in parallel.

For campaign managers: before using AI for audience or budget decisions, ask whether you can form your own hypothesis about a performance problem before looking at the summary. If not, work from the raw data until you can. Then use AI to move faster on what you already understand.

The AI Council's push for adoption is correct. The gains go to the people who already know what good looks like.

Common questions