Artificial intelligence is advancing in companies at a speed greater than the capacity of many organizations to govern it. This technology has already entered the daily lives of teams, but its accelerated adoption is exposing an increasingly evident gap. Many companies try AI, but few manage to scale it, measure its return or adequately control its risks.
According to data analyzed by technology consulting firm h&k, 88% of organizations use AI in at least one function, but only 33% have begun to scale their pilots to production. At the same time, 79% do not have a mature governance model for AI, only 25% of initiatives deliver the expected return, and only 20% of companies use these tools to increase revenue. For the technology consultancy, these data show that the real challenge is no longer in accessing AI, but rather converting it into a governed, secure and business-connected capability.
“AI is reaching SMEs before their control models. Many organizations have started by testing tools, but have not yet defined who governs them, what data they use, what risks they assume or how they are going to measure their impact. The risk is not in using AI, but in scaling it without a clear strategy,” says Javier Tejada, co-president and head of technology at h&k.
Ten key lessons for AI deployment
Given this scenario, h&k has identified ten key lessons to help SMEs apply AI in a responsible, sustainable way and in line with new European regulatory requirements.
1. Access does not guarantee adoption. Having licenses or training teams in a timely manner does not ensure real use. 60% of those who have access to AI use it daily, but only 36% feel well trained. This difference shows that the problem is not only technological, but organizational. Adoption should be measured by the change in habits, active use and integration of AI into the real tasks of each profile.
2. The bottleneck is moving from pilot to production. Many companies have already tried AI, but are unable to bring these projects to the regular operation of the business. Only 33% have begun to scale to production and 74% recognize difficulties in
generate value beyond the pilots. The consulting firm h&k warns that launching tests without success criteria or data increases the risk of accumulating projects that never leave the laboratory.
3. AI is being used more to save than to grow. 66% of organizations use AI to improve productivity and 40% to reduce costs, but only 20% apply it with the aim of increasing revenue. Without a doubt, this trend limits its strategic potential. Efficiency may be a good starting point, but competitive advantage will come when AI is connected to growth processes.
4. Automating poorly designed processes does not solve the problem. One of the most common mistakes is applying AI to processes that were already inefficient. According to h&k, adding artificial intelligence to unclear, redundant or poorly documented workflows generates limited value. Before automating, companies must review their processes, organize responsibilities, prepare data and redesign workflows where AI can add real value.
5. Governance is late, especially with AI agents. 79% of organizations do not have a mature governance model for AI. This lack of control becomes more critical with the emergence of agents capable of executing actions, making decisions or connecting with corporate systems. h&k recommends establishing an AI committee, human oversight, and clear criteria for what each solution can do from the start.
6. Many companies do not know how to measure return. Only 25% of AI initiatives deliver the expected ROI and only 6% achieve an EBIT impact equal to or greater than 5%. Part of the problem is measuring activity instead of results. It’s not enough to know how many people use a tool or how many pilots have launched; But you have to measure real savings, incremental income, reduction in time or changes in productivity.
7. The dispersion of use cases can slow down the transformation. The pressure to “do something with AI” is leading many companies to accumulate dozens of ideas without clear prioritization. h&k recommends avoiding endless lists of pilots and starting with fewer than 15 high-value cases, with managerial drive, technical feasibility and concrete metrics. And the excess of pilots consumes more resources than it releases.
8. A chatbot is not an autonomous agent. 62% of organizations experiment with AI agents, but only 23% manage to scale them. Additionally, only 10% of organizations with agentic AI report significant ROI. The leap from a chatbot to an autonomous agent is much bigger than it seems; since it requires clean data, governed APIs, standardized processes, risk limits and exhaustive recording of decisions.
91% of companies plan to increase their investment in AI in 2026, despite the fact that many have not yet obtained the expected return
9. The labor challenge is not only to replace, but to requalify. 31% of the workforce will need retraining in the next three years. Therefore, considering AI only as a way to reduce staff can generate rejection, uncertainty and poor adoption. The key is to train teams, redesign functions and move talent towards higher value activities.
10. AI requires sustained investment, not isolated impulses. 91% of companies plan to increase their investment in AI in 2026, despite the fact that many have not yet obtained the expected return. For h&k, this apparent contradiction reflects that AI must be understood as a medium-term transformation that combines first visible results with transformative projects.
From these lessons, h&k summarizes the decisions that separate leading organizations from the rest in four: 1-ambition to treat AI as a business transformation, 2-approach to prioritize fewer cases, but better executed, 3-discipline to implement governance, KPIs and periodic reviews and 4-staff training, to accompany adoption with training, requalification and redesign of roles.
Deploy the AI Master Plan
In this sense, the company defends that artificial intelligence only generates value when it is integrated into the corporate strategy, connects with key processes and has a governance model that guarantees security, scalability and real impact on the business. To achieve this, h&k proposes working from an AI Master Plan that helps define which use cases to prioritize, which enablers to build first, and how to scale the technology in an orderly manner.
As a first step, h&k has made its AI Self-Diagnosis Guide available to companies, a resource designed to help organizations identify where they are, what barriers are blocking their initiatives and what the next step should be to apply artificial intelligence with real impact. The AI Self-Diagnosis Guide is available for download.
