2025 ended with a paradox that organisations can no longer afford to overlook: never have we seen so many AI initiatives deployed, yet never has the gap between adoption and real business value been so clear.
According to McKinsey’s latest report, The State of AI in 2025, 88% of companies already use AI in at least one business function, but only a third have managed to scale beyond the experimental phase. This suggests that while technology has entered organisations, those organisations are still not designed to coexist with it. This disconnect explains why so many initiatives fail to progress. The problem does not lie in the model itself, but in the underlying structure that should support it.
What is particularly striking is that the companies achieving tangible benefits behave fundamentally differently. Known as “AI high performers”, this small group, representing only 6% of companies, has generated up to 5% of its EBIT through AI-driven initiatives. They share a common trait: they do not focus on scaling models but on scaling decisions. Rather than investing in isolated tools, they build ecosystems that align strategy, data management and governance. Their advantage is so pronounced that high performers are almost three times more likely to have re-engineered entire workflows to integrate AI. While most organisations are still trying to adapt and understand their processes, these companies are already reorganising how decisions are made.
Meanwhile, research from MIT Sloan in collaboration with BCG, The Emerging Agentic Enterprise, highlights an even more revealing trend. Generative AI adoption reached 70% in only three years, and autonomous agents reached 35% in just two years, with a further 44% planning deployment. Technology is advancing at a pace that organisations are structurally unprepared to absorb.
This reality creates a structural tension that will shape 2026. Agents are no longer an experiment; they are becoming a new organisational layer. As companies begin to delegate complete stages of a process to autonomous systems, governance stops being optional and becomes indispensable.
Semantic Governance
Traditional data governance, long centred on compliance and standardisation, is proving insufficient in a landscape where models can learn, plan and execute autonomously. A significant number of organisations have already experienced incidents resulting from poorly governed AI, with inaccurate outputs among the most common causes.
Even more concerning is the gap between the importance attributed to explainability, meaning the ability to understand which data is used, how it is transformed and under what rules, and the actual measures organisations have implemented to manage it. The data lifecycle is still perceived as a black box, even in essential areas such as traceability and lineage. This undermines trust and exposes the fragility of traditional governance approaches, which are no longer adequate to sustain strategic investment in AI and data at scale.
At Keepler, we see an imminent evolution in governance. Operationally, it will no longer function as a static set of policies. Instead, it will become an active mechanism that intervenes precisely where decisions are made and that supports models, agents and teams at the exact point where impact occurs.
However, governing decisions requires something deeper: meaning. This is where data management reaches an important inflection point. For many years, the discipline revolved around architectures, pipelines and standardisation. The expansion of AI now demands a shared layer of language that enables systems to interpret the business accurately. Organisations must understand, both technically and functionally, what each asset represents, what rules govern its use, how domains relate to one another and which semantic variations are acceptable.
The shift towards ontologies, knowledge graphs and cognitive data products is increasingly present, not as a theoretical aspiration, but as an operational requirement that allows technology, people and business to reason within the same context. In this environment, governance will not focus solely on data. It will ensure that business conditions are faithfully represented in the use cases being implemented.
At Keepler, we help organisations identify that real value emerges when data is prepared for AI in a consistent, semantic and governable manner. Evidence from the market shows that companies with higher data maturity achieve improvements of up to 20% in revenue, cost efficiency, risk reduction and productivity. As this transition continues, data strategy reasserts its evolution by shifting from a conceptual compass to the operating system that connects business objectives with technical capabilities.
The Economics of AI
Another critical dimension is prioritisation based on economic impact. Companies are no longer willing to invest without a clear estimate of return, and this expectation is evident in the organisations we work with. The ambition is clear: they want to capture the potential value of their data but often lack a coherent and cohesive portfolio.
The conversation has shifted from the question of “what can we automate?” to the more strategic “which decisions genuinely influence our business objectives, and what is the value of improving them?”
Market analyses estimate AI’s global economic potential at between 17.1 and 25.6 trillion dollars. Although this helps illustrate the scale of the opportunity, many organisations still struggle to capture it. The reason is simpler and more uncomfortable: most continue prioritising technical use cases rather than decisions with true business value.
This is where Keepler’s Navigator methodology becomes truly differentiating. It does not focus on evaluating AI and Data Initiatives. Instead, it structures enterprise reasoning around business value. Rather than generating long lists of initiatives with vague technical appeal, Navigator requires technical and business teams to jointly answer questions that are often avoided.
- What specific decision are we trying to improve?
- What is its economic impact today?
- What technical and organisational dependencies exist?
- What is the cost of doing nothing?
When portfolios are organised through this lens, roadmaps change. Non-scalable efforts are eliminated and strategic decisions with realistic, evidence-based returns take priority. As a result, investment in AI and Data becomes aligned with actual business objectives rather than theoretical expectations.
Other major consultancies are reaching similar conclusions. Bain, for example, reports EBITDA improvements of 10% to 25% in companies that have successfully integrated AI into their operating model. Additional studies highlight how well-governed agents can automate up to 80% of tasks and reduce manual handoffs, generating tangible operational benefits
Our work with organisations in complex environments has shown that meaningful results only emerge when a solid structure exists to support autonomy. Deploying technology in isolation is not enough. Such a structure rarely comes predefined; it must be built, organised and, above all, governed.
This is why, at Keepler, we see the conversation about AI and Data shifting. It is no longer about technical teams identifying what a model can do. Instead, organisations must collectively understand how to make decisions with economic rigour. Value stops being an emergent outcome and becomes an explicit mechanism that must be designed, prioritised and measured. When this logic permeates data strategy, management and governance, AI stops behaving like an isolated experiment and becomes a tool that transforms the economic structure of the organisation in a sustainable way.
A Systemic Approach
The dynamics anticipated for 2026 reveal a structural shift in how companies integrate technologies such as AI into their operating model. Across sectors and geographies, we observe a consistent pattern: strategy, management and data governance are no longer independent capabilities but components of a single integrated system. When these capabilities operate in silos, organisations experience friction, ambiguity and a proliferation of pilots that never scale. When they operate under a shared logic, AI becomes embedded organically and sustainably across business processes.
From our experience at Keepler, this convergence is not conceptual but practical. We have worked with organisations equipped with advanced technologies and highly skilled talent who nevertheless struggled to generate value because each discipline was pushing in a different direction. We have also seen cases where the absence of sophisticated infrastructure did not prevent AI from scaling, precisely because coherence existed across decisions, operational flows and data architecture. The real differentiator is not the technology itself, but the organisational design that supports its use.
Adoption and Human Integration
The implementation of new technologies and tools is not defined solely by alignment between strategy, data management, and data governance. It also depends on whether an organization’s talent is prepared to operate within this new ecosystem with clarity and confidence. Companies that successfully move beyond experimentation and pilot phases are not those that test the most tools, but those that build semantic frameworks that help people understand what benefits the technology offers, when it delivers value, and under what conditions it should be used.
In our experience, a data strategy becomes a true business enabler when these technical capabilities stop being perceived as something external to day-to-day work and instead become a natural extension of how employees operate. Genuine success is not possible if talent is not supported through proper guidance, upskilling, and reskilling across the different phases of technological renewal. These technologies must ultimately materialize as capabilities that allow teams to change how they analyze, decide, and execute.
Adoption ceases to be a follow-up program and becomes a direct outcome of mature data management: teams embrace new tools because data is ready, contextualized, and aligned with how the business makes decisions, and because talent management is embedded from the outset in technology build plans.
We are an Advanced Data Analytics company that makes amazing things possible by applying Public Cloud Technology and Artificial Intelligence to your data.
0 Comments