Evidence needed when AI replaces jobs

Article written for Otago Daily Times, by John Farrow.

FINANCE Minister Nicola Willis and Justice Minister Paul Goldsmith recently announced that the core public service could be brought down to about 55,000 staff. This was at a level they described as ‘‘ the historic norm of 1% of the population’’ .

The public sector is now at about 64,000 staff. The plan is to cut 8700 FTE over the next three years with a possible $2.4 billion saving. Willis also announced that Public Service Agencies will be asked to come up with proposals to merge their existing activities around public-facing functions, using common technology platforms.

The use of AI tools has featured heavily in Willis’ statements:

‘‘ AI tools will deliver better value for money.

‘‘ The government’s chief digital officer will oversee investment in digital systems to move human resources, payroll, case management and other systems to the Cloud and to imbed AI deployment as a basic expectation of all public entities.’’

‘‘ We are going to ensure government is fully on-board with the digitisation and AI revolutions sweeping the world.’’

These announcements were universally met with criticism from the Public Service Association and New Zealand Council of Trade Unions. They argue that New Zealand needs skilled and talented people to deliver the public services relied upon.

To date, the announcements have lacked detail about exactly how AI will perform tasks allowing the government to cut 8700 public sector jobs. Labour has asked what the roll-out and licensing costs of AI will be. Goldsmith has responded:

‘‘ I don’t have that exact figure at the moment but, of course, it varies.

‘‘ I’m not aware of a current local AI provider in the scale of Claude or CoPilot, but what I would say is that we’ll be making use of the best technology available.’’

Some commentators have signalled huge challenges to the government’s savings goal. Prof Alexandra Andhov, Chair of Law and Technology at the University of Auckland, has stated: ‘‘ The published material doesn’t really show the cost-side of the AI.’’

Enterprise-scale AI is not a one-off buy but has ongoing costs in licence fees, model upgrades and responding to companies dictating when and how models are replaced or integrated, plus audit and oversight of its operations.

The costs paid to AI today are heavily subsidised while the AI companies are trying to capture as much of the market as possible.

Prof Andhov believes that real-world pricing and cyber security demands ‘‘ will push up the cost of AI and generally any kind of digital infrastructure to such an extent’ ’ that he thinks no-one is yet even in a position to foresee.

Labour’s Science spokesman, Reuben Davidson, is on the record as stating that the coalition government does not seem to have an idea of what implementing AI would actually cost and if it is even capable of doing the work required.

So, what does all this mean in terms of actually implementing these job cuts and how will the Employment Relations Authority and Employment Court view such matters?

Employers’ good faith obligations to employees, when proposing restructuring and redundancies, are well-established . The employer must provide information which is relevant to the proposal and allow the employee to comment on the information before any final decision is made. Any decision will be judged by the ‘‘ Test of Justification’ ’ — were the employer’s actions, and how the employer acted, what a fair and reasonable employer could have done in all the circumstances?

The court will firstly look at the genuineness of the redundancy, and secondly, at whether the redundancy was carried out in a procedurally fair manner. If an employer can show the redundancy was genuine and that the notice and consultation requirements have been complied with, that will go a long way towards satisfying the test of justification. However, there are a number of decisions where the employer has got it wrong and suffered accordingly.

In Grace Team Accounting Ltd v Brake, Grace Team Accounting made one of its senior accountants redundant on the basis that it could not afford to keep her on. She sought the financial information relied upon to make the decision. This was initially refused.

Historically, the court has said that it will not look behind such business decisions. However, in Grace Team Accounting the court said that it needed to make sufficient inquiries to establish whether or not the actions were those of a fair and reasonable employer.

In summary, Grace Team Accounting failed to provide adequate or sufficient information. Somewhat embarrassingly, the financial analysis it conducted was found to be wrong, with the practice being significantly more profitable than was presented in the redundancy proposal. The redundancy was found to be unjustified because it was based on wrong information when the correct information was readily and reasonably available.

In another decision, the Employment Court in Totora Hills Farm v Davidson held:

‘‘ It will be insufficient under s103A (test of justification), where an employer is challenged to justify a dismissal or disadvantage in employment, for the employer simply to say that this was a genuine business decision and the court (or the authority) is not entitled to inquire into the merits of it. The court (or the authority) will need to do so to determine whether the decision, and how it was reached, were what a fair and reasonable employer would/ could have done in all the relevant circumstances.’’

The objects of the Employment Relations Act 2000 include acknowledging and addressing the inherent inequality of power in employment relationships and recognising that employment relationships must be built not only on the implied mutual obligations of trust and confidence, but also on a legislative requirement for good faith behaviour.

The court is likely to recognise that AI will genuinely justify redundancies. However, where humans are likely to lose their jobs to artificial intelligence, the court will strongly critique the data and the employers’ analysis in support of any such decision.

In the coalition government’s public sector example, it won’t be sufficient to claim savings of $2.4b over four years without also analysing the ongoing costs of AI. Similarly, it won’t be sufficient to claim that a role is redundant because AI can perform it, without providing full detail of how AI will actually perform the role. As the Grace Team Accounting decision established, a failure to provide adequate, accurate and sufficient information as part of a consultation process will mean that process is inevitably flawed.

Jeannie Paterson, professor of law and director of the Centre of AI and Digital Ethics at the University of Melbourne, believes that:

‘‘ Unless there is an investment in training and resources for humans, and a pretty clear democratic commitment to what role AI should be playing in society, then it’s sort of a rush to the lowest, to the bottom.’’

Leaders that are really capable and literate about this technology will be able to make better-informed and strategic decisions about what it means for the people working for them. It is that type of capability and analysis that will be essential to any fair redundancy process.

The counter-factual is that if those making strategic decisions aren’t sufficiently informed, any decision resulting in human redundancies in favour of AI is likely to be overturned by the country’s employment institutions.

Disclaimer: The opinions expressed in this article are those of the writer and do not purport to be specific legal or professional advice John Farrow is a partner with Anderson Lloyd, specialising in Employment Law.