Thu, 19 Nov 2026

From tasks to purpose: Why CIOs must rethink work for AI agents

Every era of work has had its own yardstick. On the factory floor, it was units off the line. In the office, it became hours at the desk, and later tickets closed or lines of code shipped. The people who produced the most, the fastest, rose to the top. That made sense when humans carried out every task themselves.

That world is changing quickly. AI agents now take on a growing share of the tasks that used to fill the working day, from pulling together requirements to drafting a first reply to a customer. As those tasks fall away, the old yardsticks say less and less about who is really adding value.

For CIOs across Asia Pacific, the question is moving from “How do we get more done?” to something harder: “What is our work actually for?” Manu Narayan, CIO at GitLab, believes the answer lies in purpose-driven work.

The ‘Future of Work’ is a decade-old idea in technology. What makes this moment different?

Manu Narayan

Manu Narayan: Every new technology shift impacts where or how we work. SaaS replaced on-premise software. The cloud replaced physical infrastructure. Collaboration tools replaced the office as the place where we get things done.

But AI is creating revolutionary change faster than any previous transformation.. AI’s capabilities to handle increasingly complex tasks in software development, for example, are doubling roughly every seven months. We’re about to see a significant shift towards non-linear productivity gains with AI redefining work itself. This comes with new questions about who, or what, gets the work done.

You argue work must shift from tasks to purpose. What does that mean for talent?

Manu Narayan: It means redefining what high performance looks like. The way we’ve organised work, measured performance, and developed talent assumed the person who produced the most, moved the fastest, and knew the most would be the highest performer. That assumption isn’t going away, but it’s no longer enough on its own.

The change can be examined through the lens of tasks and purpose. Every job has tasks. For a software engineer, coding is a task. But the purpose of software engineering is to solve problems through technology and find new challenges worth solving. In this new AI world, when tasks are continuously eliminated, purpose becomes ever more important.

The same goes for an engineer, an analyst, or a recruiter, as many of their everyday tasks can increasingly be handled by AI agents. But the purpose behind each of those roles can’t, because purpose is rooted in the value we assign to work, not execution.

That distinction can be generalised to every business function. The question is no longer “How do I do this better?” Instead, workers need to ask, “How do I direct a team of agents to handle this, so I can focus on the real value of my role?”

So, is everyone a manager now?

Manu Narayan: In short, yes. In the new human-to-agent model, we drive purpose. Everyone becomes a manager, setting intent, defining outcomes, delegating execution to agents, reviewing outputs, and course-correcting with judgement that no model can replicate.

What does non-linear productivity look like in practice?

Consider how we view productivity. I like to think of it this way: Thomas Edison didn’t solve the problem of dim candlelight by making better candles. He invented lightbulbs. Likewise, this is not about 5% or 10% productivity gains, but rather a change in the output curve itself. It’s like moving from candlelight to electricity, not from a dim candle to a bright one.

Photo by Daniil Komov from Pexels: https://www.pexels.com/photo/close-up-of-computer-screen-with-code-and-menu-options-34804017/

Picture a software engineer getting ready to ship a new feature. Today, they bounce between a project tracker for requirements, a wiki for architecture decisions, a code repository for recent changes, and messaging threads for open questions from product. The ramp-up alone can take hours, and it repeats every sprint.

Now hand that work to a team of agents running in parallel, pulling requirements, flagging architectural conflicts, summarising recent changes, scanning for vulnerabilities, and drafting an initial implementation plan. The engineer reviews and gets straight to work.

This approach carries across functions:

  • A support team cuts time-to-resolution by 50% because agents assemble context and draft responses before a human touches the ticket.
  • A legal team reviews a contract in minutes instead of days, with a research agent pulling precedents, a compliance agent flagging risk clauses, and a drafting agent proposing redlines, leaving the lawyer to focus on negotiation strategy and judgement calls that require human expertise.

Are enterprises seeing these gains yet?

Manu Narayan: It’s already happening. EY recently announced that nearly all enterprise organisations they surveyed reported AI productivity gains, with about half saying they were significant, and that those gains were reinvested for growth, not headcount reduction. That momentum is real, but it also raises the stakes: the biggest risk in AI adoption right now is moving too slowly.

APAC is full of AI-native challengers. How exposed are incumbents?

Manu Narayan: More than many realise. AI-native companies are already shipping products faster and without the process overhead most enterprise organisations carry.

For startups, some of that advantage is a function of size. But a meaningful portion is structural: they started building with AI at their core out of necessity, not as an incremental layer on top. They didn’t have to retrofit because AI was a partner from the start.

We see companies across the region challenging incumbents, from Singapore’s digital banks to Southeast Asian e-commerce platforms. That means the operational lag that places many larger organisations at a disadvantage is getting an order of magnitude worse.

What separates the organisations that will pull ahead?

Manu Narayan: It won’t be the number of pilots they run. The organisations that pull ahead will build the underlying structure that lets agents act with speed and safety. That includes clear ownership of AI decisions, shared context, and guardrails that scale. Without that, adoption fragments into silos and pulls people into experimentation that distracts from the work that actually matters.

In Singapore, IMDA’s Model AI Governance Framework for Agentic AI puts human accountability at the centre of agent deployment.

What should CIOs in Asia prioritise in 2027?

Manu Narayan: Unlearning a lot of the preconceived assumptions we have about work. For enterprise organisations, this will be much more challenging given the pace at which AI is advancing.

The window is closing, but we can’t just move faster; we need to move differently. As purpose becomes more essential, the judgement and intuition that make us all human have never been more valuable. The organisations that capture AI’s non-linear gains will understand the purpose of their work better than anyone.

Related:  ExecOpinion: Strategies for women building a cybersecurity career

Related Stories

MORE STORIES

Subscribe