AI Is Giving Us More Reasons to Hire

AI is already creating work. Over the short to medium term, I expect the larger employment gain to come as businesses find customers they could not previously afford to serve. A company looking only for staff savings may miss the best part of the opportunity.

Over the short to medium term, I expect AI to create more jobs than it destroys. Both blue collar and white collar jobs.

That may sound a strange view after another round of redundancy announcements, but I’m not so sure all of that is due to AI. It’s due to other fiscal related matters. That doesn’t mean the company making the redundancies will admit that. Using AI (AI washing) as an excuse for the redundancies sounds much better for shareholder sentiment.

Make expertise more affordable and a service that was previously too expensive to provide can become a viable business. That doesn’t mean salaries need to drop. I mean lowering the barrier to accessing expertise, so a company can do things it previously couldn’t afford to do.

The difficulty is that job losses are easier to count than the jobs created this way. A company announces 500 redundancies and the number is immediately visible. A service becomes viable at a lower price, finds customers over several years and gradually employs more people. Its contribution is much harder to identify as an AI employment story.

We do not yet have a reliable economy wide count proving a net gain from AI. My view is about where this leads for the short and medium term. The longer term is much harder to predict. By longer term I mean a decade out. The current hiring evidence, employers’ expectations and the economics of serving a larger market give me reasons to be optimistic. They also suggest that management has more influence over the outcome than a forecast of inevitable job losses allows.

Some of the new work needs a toolbox

Indeed’s US data show data centre job postings more than doubled between June 2024 and June 2026, while total postings fell roughly 12%. About a quarter of the data centre openings were for installation and maintenance. Among positions advertising hourly pay, the median was 42% above comparable roles outside data centres. These are advertised opportunities and pay rates, rather than a count of completed hires. ^[1]

Those buildings need electrical systems, cooling and people who know how to keep the equipment operating. The employment extends well beyond software development. For someone working in a trade, the immediate AI opportunity may be much more physical than the public debate suggests.

Construction jobs and operating jobs have different lifetimes. A building project finishes; maintenance continues while a facility runs. Both depend on the economics of the investment. But the demand for that labour is visible now, during a period when broader US job postings have weakened.

I would not base a lasting employment argument on a construction boom alone. The more interesting stage comes when the infrastructure supports services that businesses and households want to keep buying. That is where lower costs can expand the amount of work available across many industries.

Employers are planning for both creation and displacement

The World Economic Forum’s 2025 employer survey projects roughly 11 million jobs created and 9 million displaced globally by AI and information processing technologies over 2025–2030. The category is broader than generative AI, and the figures are rounded employer based forecasts. They are not a tally of jobs already gained. ^[2]

What I find useful is that employers considering both effects expect creation to exceed displacement. The people making investment and hiring plans do not see only a shrinking need for labour.

The forecast will not describe every business or occupation. Some companies will reduce staff. A growing company may employ fewer people in one function and more in another. Even a positive total can involve difficult changes for the people whose existing work disappears.

A forecast also tells us less than the commercial mechanism behind it. I put more weight on whether AI makes additional work worth paying for. There is no shortage of needs. The constraint is often the price of meeting them.

Lower costs can open a market

Consider a specialist service that works economically for a large corporate customer but is too expensive for a small business. Both customers may have the same underlying need. The smaller one simply cannot justify enough of an expert’s time at the existing price.

If AI reduces the time needed to prepare the analysis, review material or produce a first draft, the provider may be able to offer a smaller service profitably. People still have to understand the customer, check the work and take responsibility for the advice. Serving each customer requires fewer hours, but the number of customers the business can reach may grow substantially.

There is a simple calculation behind this. Suppose a service currently requires 100 hours of human work to deliver a particular volume. With AI it takes 80 hours, including the checking and correction that remain necessary. Labour per unit has fallen by 20%.

If output stays the same, the business needs fewer hours. If output grows by 25%, it needs the original 100 hours. At 40% more output, it needs 112 hours. The same improvement in the tool can produce any of those outcomes, depending on demand.

The illustration assumes quality and productivity hold as volume increases. Hours are not headcount, and the AI service, computing and implementation have costs of their own. A reduction in human effort makes the service cheaper overall only if the saving exceeds those costs.

Even with those qualifications, it is a serious omission to estimate employment from hours saved while assuming the market remains unchanged. Price affects what people buy. A business may purchase analysis it previously went without, offer more individual attention to its customers or finally deal with a backlog that was too expensive to address.

Not every saving will generate enough demand to support more jobs. Some markets are already well served; others have limited scope to grow. But treating those cases as the only possible outcome leaves a substantial commercial opportunity unexplored.

People can become productive sooner

Erik Brynjolfsson, Danielle Li and Lindsey Raymond studied an AI assistant used by 5,179 customer support agents. Their 2023 working paper reported an average 14% improvement in issues resolved per hour, with larger gains among less experienced workers. ^[3]

That result gives a manager options. A service losing customers to long queues might answer sooner and retain more of them. Another might take on additional customers without increasing staff at the previous rate. Whether employment then grows depends on how much extra business follows.

The result for less experienced staff is especially interesting. Where a tool helps a new employee perform useful work sooner, it can improve the economics of recruiting and training. Knowledge that previously depended on finding an experienced colleague becomes easier to draw on during the work itself.

Someone still has to teach the service and review mistakes. Faster answers are of little value if customers cannot rely on them. But it is possible to invest in a tool and become more willing to employ people because of it. That possibility gets too little attention when a productivity improvement is presented as a redundancy calculation.

The business case should include new customers

If two firms achieve the same saving per customer, they need not make the same decision. One may reduce staff because it has no further demand to serve. The other may use some of the saving to reach a group of customers that was previously unprofitable.

I would want to see that second option examined before accepting a proposal built entirely around fewer employees. Who could now afford the service? What would they buy, and would the business earn enough at that price to make expansion worthwhile?

Those questions can be tested. Offer a limited service to a defined group of customers and measure whether they use it, return and pay enough to cover delivery. Include the time spent correcting errors. If the economics work, the business has a reason to expand and a much clearer idea of the people it needs.

There is no obligation to preserve tasks that customers no longer value. Equally, reducing a payroll is a narrow measure of success for a technology that might give the company access to a larger market. I would rather know whether the business could grow faster than its labour requirement per customer is falling.

For Australian organisations, the US hiring figures and global survey provide useful context. Local wages, available skills and customer demand will determine the result here. An actual service that people keep buying is more convincing than a general promise about what AI will do for the economy.

More jobs overall would not make the transition easy

There are warning signs for younger workers. Stanford’s August 2026 update, using ADP payroll data through June, found employment among people aged 22 to 25 in AI exposed occupations 19% below the level implied by keeping pace with less exposed peers. Weaker hiring was the main channel. The researchers describe an early descriptive signal, rather than a causal estimate, and found no widespread economy wide displacement. ^[4]

This is a key point that I want to focus on alongside the growth case. A new technical vacancy is not automatically an opportunity an administrative worker can take. Geography, training and the time needed to change occupations all matter. A positive total would offer little immediate comfort to someone facing those barriers.

Employers also need to consider how people acquire judgement. Experienced staff often learnt through the routine work now being automated. If a business removes the entry path without replacing the learning it provided, it can make its own future hiring problem worse.

Supervised practice should be part of the new roles. Give people work they can gradually take responsibility for and access to experienced colleagues who can explain mistakes. The productivity gains for newer workers suggest AI could help that process, provided the organisation still invests in it.

I would plan for growth

The evidence does not support a claim that every automated task removes a job. We can already see new labour demand around the infrastructure. Employers anticipate creation as well as displacement, and lower delivery costs can make services viable for customers who were previously priced out.

That is why I expect AI to create more jobs than it destroys in the short to medium term. The scale will depend on whether businesses use it to expand what they can offer, and whether people can move into the resulting work. Neither happens automatically, but both can be influenced by the decisions organisations make now.

I want an AI business case to show what the company could afford to do next, alongside the savings on what it already does. If the technology allows more customers to buy something useful, there may be a much better reason to hire than there was before. I would not want to miss that because we stopped at the cost cutting spreadsheet.