AI isn’t coming to save us

AI cannot save us, but it will facilitate a vast amount of short-term profit. The human, social and environmental costs will be considerable.

AI is being deployed as a power play by elites with everything to gain and little incentive to carry the consequences. Whether it works or, more likely, fails to deliver what has been promised, those consequences will be displaced onto everyone else: workers, communities, public institutions and the future. The gains remain concentrated; the disruption is distributed.

This is a claim about the widening gap between the authority with which AI is being deployed and the organisational capacity available to understand, govern and evaluate it. In Deloitte’s global survey of 695 board directors and senior executives, 66 per cent said their boards had limited or no knowledge or experience of AI, while 31 per cent said AI was not yet on the board agenda (Deloitte Global Boardroom Program, 2025).

The same pattern appears in Australia. The Governance Institute of Australia found that 88 per cent of respondents reported difficulties integrating AI into existing systems, while 93 per cent of respondents to the relevant survey question were unable to measure the return on investment from their AI initiatives (Governance Institute of Australia, 2025). Organisations are adopting AI while often lacking the means to determine reliably whether it is delivering value, where that value exists or what costs are being displaced elsewhere.

The contradiction is difficult to miss. AI is acquiring institutional authority inside organisations that remain unable to explain it adequately, integrate it reliably, govern its consequences or measure whether it works. IBM found that 64 per cent of surveyed CEOs acknowledged that fear of falling behind had caused them to invest in technologies before clearly understanding the value those technologies would bring. Respondents reported that only 25 per cent of AI initiatives had delivered their expected return, while just 16 per cent had been scaled across the enterprise (IBM Institute for Business Value, 2025).

This is not strategic comprehension. It is competitive imitation under conditions of institutional anxiety. Senior management is being told that failure to adopt AI is itself a failure of leadership, so investment becomes compulsory before understanding has caught up. The technology acquires authority not because it has been demonstrated to solve the problem, but because every organisation fears being the last one not using it.

I have spent a long time considering the underlying system dynamics, and one conclusion follows clearly: many of the people exercising power over this transition do not adequately understand the situation they are creating. Even optimistic industry research concedes that almost every company is investing in AI while only 1 per cent of surveyed executives regard their organisations as mature in their deployment of AI. The same research identifies leadership capability as a critical constraint on successful AI adoption (Mayer et al., 2025).

The deeper failure is not technical. Institutions remain unwilling to confront the social and cultural problems that AI is far more likely to lock in and proliferate than resolve. Systems built around inequality, administrative abstraction, precarious labour, surveillance and concentrated ownership will not be corrected by automating them. AI will reproduce the relations through which it is trained, purchased, governed and applied.

The executive class is not being rescued by AI. It is becoming trapped inside an atavistic response disguised as technological sophistication: substituting computation for judgment, scale for understanding, prediction for knowledge and automation for institutional reform. This is little more than magical thinking with infrastructure.

AI does not arrive from outside society to repair it. It enters through existing concentrations of power and becomes another means by which those arrangements reproduce themselves. Its consequences will not be determined primarily by what the technology can do, but by who controls it, what they require it to preserve, and who is forced to absorb the cost when their promises fail.

References

Deloitte Global Boardroom Program (2025) Governance of AI: A Critical Imperative for Today’s Boards. 2nd edn. Deloitte Global. Available at: Deloitte website (Accessed: 2 August 2026).

Governance Institute of Australia (2025) 2025 AI Deployment and Governance Survey Report. Sydney: Governance Institute of Australia. Available at: Governance Institute of Australia website (Accessed: 2 August 2026).

IBM Institute for Business Value (2025) 2025 CEO Study: 5 Mindshifts to Supercharge Business Growth. Armonk, NY: IBM. Available at: IBM website (Accessed: 2 August 2026).

Mayer, H., Yee, L., Chui, M. and Roberts, R. (2025) Superagency in the Workplace: Empowering People to Unlock AI’s Full Potential. McKinsey & Company, 28 January. Available at: McKinsey & Company website (Accessed: 2 August 2026).