Models Don’t Transform Organisations. Decisions Do.

By Ignacio Barahona, Founding Partner and Head of Data Analytics & Modern BI at Innova-tsn

Over the past few years, many organisations have measured the success of their data initiatives by the sophistication of their models, the accuracy of their algorithms or the capabilities of their platforms. Yet the real question should be a different one: is that data actually being used to make better decisions? In many cases, the answer reveals that the main challenge is not technological, but organisational.

Today, the issue is no longer simply about building models, dashboards or advanced analytics systems. The real challenge is ensuring that these solutions fit into day-to-day operations, address genuine business needs and are embraced by the people expected to use them. A technically brilliant solution can still fail if it enters the process too late, does not integrate with existing systems, fails to account for regulatory constraints, or if the end user does not understand why they should trust it.

As a result, many initiatives fall by the wayside because they are conceived from an overly technology-driven perspective, becoming enamoured with the solution before the problem has been properly validated. When this happens, the outcome is often familiar: use cases that never make it into production, investments that fail to scale, or tools that, despite working correctly, fail to deliver any meaningful impact.

Adoption Starts Before Deployment: The Importance of Data Governance

Adoption cannot be treated as the final stage of a project. It is not something that begins once the model has been trained or the dashboard has been designed. It starts much earlier: with defining the use case, engaging with the relevant business areas, understanding the actual process and identifying the people who will interact with the solution in their day-to-day work.

This is where the role of the business becomes critical. Business users cannot simply be recipients of data products; they must be involved in their design, validation and evolution. When users are engaged from the outset, they provide context, anticipate potential friction and build trust. Without trust, there is no adoption — and certainly no willingness to delegate decisions.

In this context, data governance ceases to be a bureaucratic layer and becomes an enabler. Governance does not mean holding back innovation; it means creating the conditions for innovation to scale with the right safeguards in place.

Data quality, security, explainability and accountability are essential if the business is to trust the solutions made available to it. However, governance does not mean controlling everything with the same level of intensity. The key is to prioritise the data and models that actually go into production, affect critical processes or contribute directly to decision-making.

Good data governance should work like a safe runway: it enables organisations to move faster by building confidence, not by creating more obstacles.

From Closed Projects to Living Products

We also need to move beyond the mindset of projects that are delivered and then forgotten. Data products should be managed as living assets, subject to ongoing monitoring, continuous improvement and measurement.

Technical accuracy matters, but it cannot be the only measure of success. A model may be highly accurate and still deliver limited value if it is rarely used or does not influence any meaningful decisions. Organisations should therefore incorporate metrics for usage, adoption, satisfaction and business impact from the outset.

This approach requires a shift in mindset. Data teams are no longer simply providers of reports or developers of models; they become facilitators of products that solve specific problems, integrate into real business processes and evolve alongside business needs.

This shift will become even more important with the rise of generative artificial intelligence and, in particular, agentic AI. Until now, the challenge has been to ensure that people use data, reports or algorithmic recommendations. The next step will be preparing organisations for that data to be interpreted and used by intelligent agents as well.

This requires us to rethink many of the foundations of data strategies. It is no longer simply about designing intuitive products for human users, but about creating platforms, semantic models and governance frameworks capable of supporting far more distributed and automated consumption. The future user of data will not only be a person sitting in front of a dashboard. It will also be a system capable of retrieving information, understanding its context and acting upon it.

Artificial intelligence can significantly accelerate this journey, but it does not remove the need to get the fundamentals right. In fact, the further automation progresses, the more important data quality, business integration, governance and trust in models will become.

AI does not eliminate organisational challenges; it makes them more visible. Companies should therefore be asking themselves not only whether they have the capability to develop advanced models, but whether they are ready for those models to be used, understood and governed — and ultimately to generate value.

This is precisely what distinguishes an organisation that experiments with data from one that is truly data-driven: its ability to make data a natural part of decision-making. The value of artificial intelligence will not lie solely in the power of its models, but in the maturity of the organisations capable of integrating them into their processes, governing them effectively and building the trust required for people to use them.

Ultimately, data does not transform organisations simply by existing; it transforms them when it becomes action.