The productivity dividend: No one should waste time on tasks that AI can do better

By Richard Sebaggala

 

A quiet revolution is underway in the UK civil service. More than 400,000 civil servants are being trained to integrate artificial intelligence into their daily work. This training is not just about improving performance, but also about rethinking what work should look like in the first place. The policy mantra underpinning this change is as direct as it is radical: "No person's substantive time should be spent on a task where digital or AI can do it better, quicker, and to the same high quality and standard digitally or through artificial intelligence."

 

It's not about replacing people. It's about distributing work more intelligently. It's about realising that not all hard work is productive and that not all productivity has to come at the expense of people. What we're seeing is the emergence of a productivity dividend; a return on investment that comes from considering how time is used and who or what is best suited to the task at hand.

 

For those of us who have called on students, researchers and institutions in Africa to engage early with AI, this shift is both encouraging and instructive. I have often argued that the real question is not whether AI can help, but why we are still doing things manually when it already can.

 

There is more at stake here than just efficiency. This moment forces us to reckon with something deeper. As Kai-Fu Lee has pointed out in AI Superpowers (2018), previous technologies such as electricity, steam engines or the internet have enhanced human capabilities. AI is different. It competes with our cognitive abilities. It doesn't just automate, it interferes with human thinking, decision making and problem solving. This makes it a unique disruptor, but also a unique transformer.

 

The evidence is already there. In the UK, AI tools such as "Humphrey" have been used to process public consultations faster than human analysts, with equally reliable results. The government estimates that it could save 75,000 working days per year on 500 annual consultations, which equates to around £20 million in labour costs. This is not simply a budget cut. It's a reallocation of time and talent in favour of tasks that require real human insight.

 

The pace of change continues to accelerate. OpenAI co-founder Ilya Sutskever put it bluntly at a recent University of Toronto event: "Anything I can learn, anything any of you can learn, AI could do." His message was not speculative, it was a warning. The fact that AI can't do everything yet doesn't mean that it won't do it. It just means that it can't do it yet. The real danger is standing still while AI evolves past you.

 

So, why are we still teaching students to format references by hand when tools can do it in seconds? Why are policy staff reading thousands of pages of public feedback when AI can collate and summarize it more efficiently? Why do researchers spend days cleaning data when automation can do it in minutes? For those familiar with econometric analysis, refusing automation in this way is akin to insisting we should derive beta coefficients in a regression using those abstract textbook formulas we learned in our second-year econometrics class. It's like saying we should ignore powerful software like Stata or SPSS, where a simple command produces results and your real job is to interpret them and check their validity.

 

 

In Africa, where resources are limited but creativity and adaptability are abundant, we have a rare opportunity to leapfrog. We aren't burdened by the outdated and often inefficient "legacy systems" that many developed nations are trying to transition away from. Instead of having to dismantle old, established ways of working, we can design our education, government, and economic systems with AI at the center from the start. This approach allows us to see AI not as a threat to be managed, but as a powerful tool to be mastered.

 

However, this means that we can no longer hesitate. It means that students must not only be afraid of AI plagiarism, but that they must work responsibly with AI. It means pushing government departments and universities to use AI to reduce administrative backlogs so that their staff can focus on delivering meaningful services. And it means cultivating a culture where the question "Why am I still doing this manually?" becomes second nature.

 

This is not a call for blind automation. It's a call for strategic delegation. If a machine can do a task better and just as reliably, leaving it in human hands is not care, it's inefficiency.

 

The real dividend of AI is not in what it replaces. It lies in what it unlocks: our ability to think deeper, act faster, and serve better. However, this dividend will only be paid to those who choose to collect it.

The future will not wait. And neither should we.

We need to start asking ourselves, our teams, and our institutions a difficult but necessary question: Why are we still doing this manually? And if the answer is: "Because that's how it's always been done," then perhaps it's time we let AI show us a better way.