AI projects are stalling at a concerning rate. Boston Consulting Group analysis shows that three in four companies have yet to show measurable results from their AI investments. A failure rate that has been driven by a multitude of factors ranging from poor data quality, inadequate platform architecture, fragmented governance and inadequate oversight – the exact structural weaknesses that upcoming regulation will punish.
The EU AI Act will penalise businesses that lack project explainability and accountability with fines up to 7 percent of their global turnover for prohibited practices. Regulation is forcing change, and the companies that treat it as a box ticking “compliance” exercise will sit on the sidelines as competitors with appropriate governance pull further ahead.
This is a watershed moment for leadership teams, a warning to strengthen the foundations now or risk falling behind for good.
The separation is already happening
Many companies lack clarity on who has ownership of AI strategy. Compliance belongs to legal, data governance to IT, innovation to labs running pilots with no production path and risk management to teams that rarely communicate with model developers. A poor outcome is inevitable with no single executive having end-to-end accountability.
Such fragmentation is the reason why high-risk AI systems will fail to meet the compliance deadline of August 2026, as without clear ownership, risks go unmanaged. Deloitte recently refunded the Australian government after a compliance report was found to contain fabricated citations generated by AI. When no single function is accountable for verifying AI outputs, expensive mistakes are to be expected.
The problem starts in the boardroom. Deloitte research, conducted across 468 directors globally, found that only 2 per cent describe themselves as highly knowledgeable about AI. At Altimetrik, when we work with AI clients, we often ask about the levels of adoption they see in their professional lives vs. their home lives; the results are startling. Without informed governance at the top and no executive owning the outcome, technical teams are left building systems without a strategic framework or a clear risk appetite.
The companies winning this race are not waiting for regulation to force their hand. They are investing proactively in these framework elements while recognising that building governance now creates a window of opportunity – while their competitors scramble to meet the compliance deadline, they are already in place and can move with greater agility.
Infrastructure investment is the real story
Data centres are fast becoming a critical infrastructure constraint. Hyperscalers and investment companies are already responding to the computing demands of AI. The EU AI Act’s requirements for traceability and explainability cannot be satisfied without the infrastructure to monitor model behaviour in production. This means investing in cloud controls, monitoring tools, and data centre capacity before the deadline. This infrastructure investment must come first; governance frameworks cannot function without the monitoring and traceability infrastructure beneath them.
But infrastructure alone will not deliver compliance. Data must be unified across sources to create a single source of truth. Lineage must be traceable from raw input through to model output. Governance frameworks must clearly define who owns quality, who audits results, and who has the authority to intervene when systems deviate from their intended purpose. Without these initial capabilities, AI projects will not succeed.
Growth versus cost reduction
We have observed a split emerging in this area. Some companies use AI for simple cost reduction, automating tasks, replacing headcount and optimising existing processes. Others use it to grow their business, creating new revenue streams, entering new markets and building products that were previously impossible. The difference comes down to governance. Governance frameworks determine which use cases are even accessible. Without them, AI remains confined to basic automation. With it, companies can pursue the higher-risk, higher-reward applications that create new markets.
The same BCG analysis found that companies investing strategically in AI dedicate up to 64 per cent more of their IT budget to AI and expect twice the revenue increase by 2028. That budget is not going into more models. It is being set aside for infrastructure, data quality, and governance frameworks that enable AI to be deployed at scale. These forward-thinking companies have recognised that technology maturity must be matched by organisational maturity. The most sophisticated algorithms fail if the business does not know how to govern them, interpret their outputs, or intervene when the results do not align with the intended outcome.
The case for a Chief AI Officer role is therefore becoming difficult to ignore. Without executive accountability, AI strategy remains distributed across functions with no coherent direction. Boards need someone who can align innovation with risk management, translate technical capability into business value, and take ownership when things go wrong. That role does not exist in many companies – but it should.
Global alignment is coming
In October 2025, the United Nations established a Global Dialogue on AI Governance, bringing all 193 member countries into the conversation for the first time. AI’s impact is global, but its governance is fragmented – that is now changing. Companies operating across borders must prepare for a world where compliance frameworks align and where regulatory maturity in one jurisdiction creates access to others.
The urgency extends beyond regulation. AI itself is evolving faster than most governance frameworks can keep up with. The technology is moving from pattern recognition toward generating novel insights, the kind that create new markets rather than optimise existing ones. Companies without governance foundations will not be able to deploy these more powerful systems when they arrive safely. The capability gap is becoming a strategic rift.
AI and data governance is no longer a compliance function. It is the architecture of trust that determines whether AI delivers growth or automates cost reduction. The difference between companies that scale AI and companies that abandon pilots comes down to whether they got the foundations right. Most have not, but some are in the process of building them. The gap is widening, and we are moving into a future most businesses are not prepared for.
