Deloitte Just Named the Mechanism Behind Most AI Transformation Failures — and It Is in Your Budget

Cultural debt — the organisational liability that accumulates when technology investment outpaces human development — is the most useful new governance concept in AI management research this year. The data shows it is being created at scale by a 93-7 investment split that most boards have approved without examining.

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/deloitte-just-named-the-mechanism?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

Two major studies were published on the same day recently — Deloitte’s Global Human Capital Trends 2026, covering 9,000 managers and professionals across 90 countries, and IBM’s CEO Study 2026, covering 2,000 chief executives across 33 countries and 21 sectors. Together they constitute the most comprehensive empirical examination of AI’s impact on leadership and organisational structure produced to date. Both studies converge on the same conclusion: the challenge of AI is no longer primarily technological. It is organisational — and the organisations that treat it as a technology problem are systematically underperforming relative to those that treat it as a leadership and human capability problem.

The Deloitte study contains one finding that deserves more attention than it has received: 93 per cent of dedicated generative AI transformation budgets are currently absorbed by technology acquisition and infrastructure. Human capability development receives 7 per cent. Deloitte calls the consequence of that imbalance cultural debt — and defines it as the organisational liability that accumulates when technological innovation outpaces the development of the human capabilities required to direct, govern, and benefit from it.

That is not a finding about the wrong kind of technology. It is a finding about the wrong kind of investment — and the boards that approved the 93 per cent without examining the 7 per cent are the ones most likely to find themselves with AI infrastructure their organisations cannot use, capability their leaders cannot direct, and returns that never materialise.

CONTEXT AND BACKGROUND

The evidence base behind the cultural debt finding is specific and significant. Deloitte’s study found that 88 per cent of leaders recognise the importance of acting quickly on AI. Only 14 per cent believe they currently have the skills required to effectively manage human-machine interaction. That is a 74-percentage-point gap between urgency and capability — the largest awareness-to-action gap in corporate AI research to date. It is also the gap that the 93-7 investment split is designed, by its internal logic, to widen: buying more technology without proportionately developing the human capability to use it produces an organisation that is simultaneously more technically capable and less operationally effective than it was before the investment.

The IBM CEO Study adds the leadership dimension to the same diagnosis. Sixty-four per cent of CEOs globally report being comfortable making major strategic decisions based on AI-generated input. Seventy-six per cent of organisations now have a Chief AI Officer — up from 26 per cent just a year earlier, an extraordinary fifty-percentage-point increase in twelve months. Eighty-five per cent of CEOs agree that all department heads must become technology experts in their respective fields. These are signals of genuine executive engagement with AI at the governance level. They are also signals that the human capability requirement — not for AI literacy in the basic sense but for the judgment, oversight, and orchestration capability that AI-augmented leadership actually requires — is growing faster than most organisations' development investment is keeping pace with.

Gary Cohn, Vice Chairman at IBM, summarised the operating environment precisely: “The role of the CEO has always been to steer the company through periods of radical change, and what artificial intelligence is changing is the speed and impact on leadership decisions. Decision-making cycles will shorten and the boundaries between different functions will dissolve: the advantage will go to those who are able to learn, adapt and act more quickly than the competition.” That is not a description of AI as a tool leaders use. It is a description of AI as a context leaders must operate within — and it requires a fundamentally different kind of development investment than the 7 per cent currently allocated to people.

The Deloitte study provides a concrete example of what organisational adaptation looks like when it is taken seriously. Moderna, the US biotechnology multinational, merged its HR and IT functions into a single department to restructure for large-scale AI-driven growth. The rationale is precisely the point Deloitte’s research identifies as the central challenge: AI requires organisations to separate expertise from structure, organise data and technology across functions, and strengthen cross-functional accountability in ways that traditional siloed organisational design systematically prevents. Moderna’s response is not the only viable design — but it represents a deliberate answer to the question that most boards have not yet formally asked: what does our organisational structure need to look like to realise the value of the AI investment we have approved?

INSIGHT AND ANALYSIS

Cultural debt is a more precise and more useful concept than the generic change management failures that most AI transformation post-mortems produce. Change management failure is a description: the organisation did not adapt adequately to the technology it deployed. Cultural debt is a mechanism: the organisation made a series of investment decisions that systematically prioritised technology acquisition over human development, creating a compounding liability that manifests as friction, resistance, abandoned pilots, and the gap between what the AI system can do and what the organisation can instruct it to do and govern it to do safely.

The mechanism is compounding because it is self-reinforcing. An organisation that invests 93 per cent of its generative AI transformation budget in technology and 7 per cent in people deploys AI tools that its leaders cannot fully direct. Those leaders, unable to fully direct the tools, cannot generate the outcomes that justify further human development investment. The absence of outcomes leads to either reduced investment or a continued bet on more technology — neither of which addresses the underlying capability gap. The debt accumulates. The returns do not materialise. The post-mortem attributes the failure to change management rather than to the investment structure that made adequate change management impossible.

The IBM study adds a governance paradox that compounds the cultural debt problem. Seventy-nine per cent of CEOs report decentralising decision-making processes. Eighty-three per cent simultaneously consider AI sovereignty — the ability to maintain control, transparency, and oversight over the algorithms and generative models driving business decisions — essential. Those two positions are in direct tension. Decentralising decision-making to AI agents reduces the human oversight that AI sovereignty requires. Maintaining AI sovereignty requires the centralised visibility and control that decentralised AI decision-making progressively erodes. The organisations that resolve this paradox are the ones that have invested adequately in the human judgment capability required to design, govern, and adjust the systems they are decentralising to — precisely the investment that the 7 per cent allocation does not fund.

The IBM projection for 2030 adds the most urgent planning horizon to the governance challenge. Senior executives expect to entrust almost half of operational decisions with codifiable criteria entirely to AI agents without any human intervention within four years. That is a structural reorganisation of organisational decision-making that will happen within the current board term of most directors sitting in South African boardrooms right now. The boards that design their governance frameworks for that reality now — that define what human oversight means when half of operational decisions are machine-made, what accountability structures apply to AI-agent errors, and what controls ensure that codifiable criteria remain aligned with the organisation’s values and obligations — will be better positioned than those that encounter that reality without having prepared for it.

IMPLICATIONS

For South African boards and executives, the Deloitte and IBM findings carry three specific implications that most current AI governance frameworks have not yet addressed.

The first is about investment rebalancing — and in the South African context, this is not merely an operational preference. It is a governance mandate. Under the King IV Report on Corporate Governance for South Africa, Principle 12 places an explicit fiduciary duty on governing bodies to ensure that technology governance aligns with risk management and organisational capability. The board is required to approve policy and strategy that leverages technology to improve organisational performance, while ensuring risk is managed and ethical culture is maintained. A board that approves a 93-7 generative AI budget split is funding the technology asset while starving the internal control environment — the human judgment, governance design, and capability development — required to fulfil that obligation. Cultural debt, in the King IV framework, is not merely a strategic risk. It is a potential failure of fiduciary duty. Correcting it requires boards to ensure that AI transformation business cases include a proportionate human development component as a governance requirement, not as an afterthought to the technology procurement.

The second is about the Chief AI Officer role. The IBM finding that 76 per cent of organisations now have a CAIO — up from 26 per cent a year earlier — represents either a genuine governance advance or a title without authority, depending on whether the role has been designed with the mandate, the board access, and the cross-functional accountability required to actually govern AI at enterprise scale. South African organisations that have appointed a CAIO without examining those questions have created an organisational response to a governance requirement without necessarily creating the governance capability the requirement demands. The role is meaningless without the investment — in both the technology and the human capability required to exercise AI sovereignty — to make it effective. Under King IV, the governing body is ultimately accountable for technology governance, which means the CAIO’s mandate must be clearly defined in terms of the board’s own accountability framework.

The third is about the 2030 planning horizon. IBM’s projection that half of operational decisions will be entrusted to AI agents without human intervention by 2030 is not a distant scenario. It is a governance planning obligation for boards whose current terms extend into that period. The accountability frameworks, the AI sovereignty mechanisms, the human oversight structures, and the organisational design required to operate effectively in that environment need to be in development now — not because the technology requires it, but because the human capability to govern it takes longer to build than the technology takes to deploy. That asymmetry — technology faster than governance — is precisely what the cultural debt concept names, and precisely what the 93-7 investment split is producing at scale. Organisations that redesign technology, finance, HR, operations, and cross-functional collaboration in a coordinated manner are, according to the IBM study, up to four times more likely to achieve their business objectives. That is not a soft benefit. It is an empirically measured performance advantage — and it is available to South African organisations that choose to invest in it.

CLOSING TAKEAWAY

Deloitte’s cultural debt concept will not appear in most AI transformation business cases. Neither will the 74-percentage-point gap between the 88 per cent of leaders who feel urgency and the 14 per cent who feel capable. Neither will the 93-7 investment split that is producing both. These are the variables that determine whether the AI transformation investment a board approves will produce the returns it projects — and they are systematically absent from the analysis that precedes the approval.

For South African boards, the obligation is both commercial and statutory. King IV Principle 12 requires governing bodies to ensure that technology serves the organisation’s performance while managing risk and maintaining ethical culture. Cultural debt — the compounding liability of underfunding the human side of AI transformation — is the mechanism by which boards fail that obligation without ever intending to. The organisations that recognise this and rebalance their investment accordingly will not only outperform their peers on AI returns. They will be fulfilling the governance mandate that their fiduciary role has always required.

Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net