AI risk #21/99: The AI saving in your plan needs a 90-day check

In February, WiseTech Global said it would cut about 2,000 jobs over two years, close to 29% of its staff. Its chief executive told the market that "the era of manually writing code as a core act of engineering is over" and that the company would run on "a structurally lower cost base". The share price rose on the day.

That second quote is a business objective, with AI expected to deliver it. Plenty of Australian and UK businesses have put a version of the same line in their plans this year, usually with less fanfare. Some of the ones that acted on it are already hiring the role back.

What's happening in Australia and the UK

Nearly 4,500 Australian tech workers were laid off in the first three months of 2026, and AI was cited as a reason for every one of those cuts, according to the ACS's Information Age. Commonwealth Bank cut hundreds of chat-support roles in July as it extended AI across customer service. The same report notes that in August last year the bank announced 45 cuts for the same reason, then reversed them and said the announcement had been made in error.

In the UK, the Financial Times reported on 7 October that HSBC is cutting about 70% of the financial advisers in its UK wealth business, and about half of the managers and specialists, as it moves to serving wealthy clients with AI. HSBC hasn't confirmed the figures, and the people affected are due to leave at the end of this month. Three years ago the bank's plan for the same business was to add more than 100 wealth managers on the way to £100bn of UK assets by 2030.

How often the cut is reversed

Recruiter Robert Half surveyed hiring managers in the US and Canada this year. Just under one in three of those who had cut a role mainly because of AI later rehired for the same or a similar role. The most common reasons were that AI needed more human oversight and quality control than expected, and that the business still needed people's relationship skills and knowledge of how things work.

Nicole Gorton, a director at Robert Half, told Information Age that for Australian employers "the biggest impact may be a shift from broad assumptions about what AI can replace towards a more considered assessment of where technology adds value and where human capability remains critical".

Gartner predicted in February that by 2027 half of the companies that cut customer-service staff because of AI will rehire for similar work under different job titles. Careerminds, a US outplacement firm, surveyed 600 HR leaders who had run layoffs in the past year: 52% had rehired for eliminated roles within six months, and 8% said the restructure delivered what was promised and they'd do it the same way again. In the US as a whole, Challenger, Gray & Christmas counts AI as the most-cited reason for job cuts so far this year, about one in five.

None of this points to AI emptying the job market. Australia's Department of Employment and Workplace Relations found in July "no evidence to date of broad labour market upheaval", and software developer jobs are up 25% since late 2022. The problem sits in the plan, in the assumption that AI could do the work from the day the cut was made.

Where the risk lands

A plan that gives a role to AI is usually serving a cost or capacity objective: run the business for less, or grow it without adding people. Both are sound objectives, and AI can do a lot to deliver them. The risk is in the assumption underneath, that AI could take over the whole role on day one with nobody checking. The saving gets tracked. The assumption often doesn't.

That assumption is where the AI risks are. If the cut doesn't hold, the role comes back at a new salary plus recruitment costs, and the saving may already be counted in the forecast. The people who stay may be asked to oversee AI output without the training or the time to do it well. Customers feel the gap while it's open.

Each of those shows up against the objective, as costs that don't fall or service that slips. So they belong on the risk register against that objective, next to the plan line that created them. Logged as an HR matter, they can easily miss the people who own the objective.

The legal side

In Australia, a redundancy is only genuine if the job is no longer needed and redeployment has been considered, and the High Court ruled last year that redeployment doesn't require a vacant position to exist. Law firm Kingston Reid advises employers who are consolidating tasks with AI to consider reorganising the work before calling a role redundant, and to assess the work task by task. In the UK, Acas guidance says redundancy applies when a role is no longer needed, and that employers should look at alternatives such as redeployment first. A business that advertises the same role a few months later will find that redundancy harder to defend.

What to add to the risk register

Three entries would make the assumption visible:

The lines that rely on AI. Which lines in the plan and in your AI strategy assume a role or task is now done by AI. That means the headcount assumptions, as well as the tools.

A check at 90 days. Who confirms the work is being done, and to what standard, with a date and an owner.

The cost of reversal. What it would cost if the role comes back, including salary, hiring and any drop in service in between. Put it next to the expected saving so the two can be compared at the check.

The AI Disruption Index shows which of your objectives rest on an assumption like this one. It starts from your objectives, scores and ranks the AI risks against each, and drafts a safeguard for every risk. It's free, takes about ten minutes, and there's a sample report on the page if you'd like to see the output first.