Eighty-eight per cent of companies are actively deploying artificial intelligence. Fewer than twenty per cent have seen any significant, tangible impact on their operations.
That figure, from McKinsey's State of Organizations 2026, should stop every CEO in their tracks. Not because AI does not work. It does, in the right conditions. But because the gap between investment and return is not a technology problem. It is a leadership, strategy and governance problem.
Deloitte's 2026 State of AI in the Enterprise confirms the picture. Only one in three organisations is using AI to deeply transform its processes or business model. The remaining two thirds are either experimenting at the margins or, as Deloitte puts it, "using AI at surface level, with little or no change to existing processes."
If your business is spending on AI but not seeing meaningful returns, you are in the majority. The question is what to do about it.
Why the Gap Exists
The failure is rarely about the tools. It is about the conditions in which the tools are deployed.
Five structural reasons explain why most AI investments fail to generate measurable business value:
- No clear problem definition. Many businesses acquire AI tools before identifying the specific operational or commercial problem they are solving. Tools get layered onto existing workflows. The workflow does not change. The outcome does not change.
- Poor data foundations. AI systems are only as reliable as the data they run on. Fragmented data, inconsistent data entry practices, and ungoverned data flows mean the model produces outputs that cannot be trusted. IBM's research suggests companies with strong data integration achieve nearly three times the AI ROI of those with fragmented data infrastructure.
- Absent executive ownership. When AI projects are delegated entirely to IT or technology teams, they rarely scale. Deloitte's research found that organisations where senior leadership actively shapes AI governance generate significantly greater business value than those where governance is treated as a technical matter.
- Measurement is an afterthought. Forty-four per cent of executives say generative AI is the hardest technology for which to define and measure ROI. Without pre-agreed success metrics, AI projects drift. Benefits get described in vague terms. Nobody can demonstrate whether the investment is working.
- Change management is underestimated. Research by MIT found that 31% of employees admit to actively undermining company AI efforts, whether through refusing tools, entering poor-quality data, or slow-rolling adoption. Deploying a tool without managing the human change around it is one of the most consistent reasons implementations stall.
What the 20% Do Differently
The organisations seeing meaningful AI returns are not necessarily the best-funded or the most technically sophisticated. They share a set of strategic disciplines that the majority overlook.
They start with a business problem, not a technology purchase. Before selecting any tool, the highest-performing organisations define a specific operational or commercial challenge, set a measurable target, and then evaluate AI options against that target. Technology follows strategy. It does not lead it.
They fix the data before they deploy. Rather than accepting that the data will "sort itself out," leading organisations conduct a data readiness assessment before implementation. They identify gaps, inconsistencies and governance failures first. This is unglamorous work. It is also the foundation of everything that follows.
They appoint a senior internal owner. Not a project manager. A senior leader with decision-making authority, cross-functional credibility, and the mandate to drive adoption. This person bridges the technical and commercial sides of the business. In smaller organisations, this is often the CEO or COO directly.
They define success in commercial terms. Not "percentage of users trained" or "tools deployed." Instead: cost reduced by a specific amount, time-to-close shortened by a measurable number of days, customer response time cut to a defined target. If the success metric cannot be expressed in business outcomes, the implementation lacks a foundation.
They treat change management as core infrastructure. The highest-return AI deployments invest as much in the human transition as in the technology itself. Communication, training, incentives, feedback loops, and leadership modelling all matter. People do not resist AI because they are backwards. They resist it when they do not understand the purpose, do not trust the output, or fear the consequence for their role.
The AI Execution Diagnostic
Before making any further investment in AI, every leadership team should be able to answer five questions honestly:
- What specific business problem are we solving?
Can you state it in one sentence? Can every member of your leadership team state it the same way? If not, the implementation lacks a clear mandate.
- Is our data fit for purpose?
Where does the data come from? How is it governed? Who is responsible for its quality? If you cannot answer these questions confidently, data failure will undermine the deployment.
- Who is accountable?
Is there a named senior leader responsible for both the implementation and the business outcome? Is that person empowered to make decisions and drive adoption across functions?
- How will we measure success?
What are the agreed metrics? What is the baseline? By when will we expect to see measurable results? If these are not documented before deployment, they will not be agreed after it.
- What is our change management plan?
How will we communicate the change to affected teams? What training is in place? How will we handle resistance? What does the six-month adoption roadmap look like?
From Experiment to Return
The AI tools available to mid-market businesses today are genuinely capable. The ceiling is not the technology. It is the leadership conditions around it.
For organisations that have run pilots but not seen results, the answer is rarely to try a different tool. It is to revisit the strategic foundations: the problem definition, the data quality, the governance structure, the success metrics, and the change management approach.
For organisations yet to make their first meaningful AI investment, the diagnostic above is the right starting point. Not a technology audit. Not a vendor comparison. A set of leadership decisions that determine whether the investment will return anything at all.
The businesses that will look back on 2026 as the year AI made a real difference to their performance are not the ones that deployed the most tools. They are the ones that deployed with the most clarity.
Allington Advisors works with founders, CEOs and leadership teams to design and implement AI strategies that deliver measurable business outcomes. If you are questioning whether your current AI investment is on the right track, or are planning your first substantive deployment, we would be glad to speak with you.