Artificial intelligence (AI) is on everyone's lips, and for small and medium-sized businesses (SMEs) and self-employed people in the technology sector, understanding its true impact is essential. Manuel Romero, co-founder and CSO of Maisa, offers a clear perspective on why so many AI projects get stuck in the testing phase, and what needs to be done to turn them into productive, reliable and traceable initiatives. In this article we explain why this approach matters and how it can make a difference in your competitiveness.
Context: Why do so many AI projects never reach production?
According to Romero, “many projects remain proof of concepts” because AI is approached as if it were magic, without asking whether it really adds value to the process. This erroneous perception leads to delegating too much responsibility to probabilistic systems without control or deep knowledge of their operation (computerworld.es).
In Spain, the adoption of AI is also uneven: while large companies take advantage of its potential, many SMEs lack the resources or technological maturity necessary to implement advanced projects (cincodias.elpais.com). Furthermore, a study in the United States reveals that only 22.8% of manufacturing industries use AI, with costs and lack of applicable use cases being the main barriers (aeaweb.org).
Practical implications: reliability, traceability and costs
Romero identifies three fundamental pillars for AI to make sense in demanding business environments:
Evaluation and return on investment (ROI): Before investing in AI, you must analyze whether it really adds value and define clear indicators to measure its economic impact (computerworld.es).
Control and support: It is not enough to delegate trust to intelligent models; Control mechanisms, contingencies and a vision of AI as a “dumb box” that needs explicit supervision must be incorporated (computerworld.es).
Traceability and auditing: Breaking the perception of the black box is key. In regulated sectors such as banking or energy, only AI is acceptable that allows each decision to be audited and ensure regulatory compliance (computerworld.es).
Maisa has built its reputation on this foundation: AI agents that execute tasks step by step, validating each decision with deterministic logic or code, reducing risks such as hallucinations (made up responses) and providing operational predictability (computerworld.es).
Recommended steps to provide value with AI
For SMEs and the self-employed to take real advantage of AI, without the risk of failure, Romero recommends a progressive and structured strategy:
Avoid excessively ambitious projects without prior analysis. Better to define two use cases per department, with identified bottlenecks (computerworld.es).
Apply AI only where it provides clear and measurable value, and include metrics from the beginning to prove its profitability (computerworld.es).
Internal training on what AI is, how it works and when it should be used: training the team avoids erroneous perceptions and allows informed decisions (computerworld.es).
Start with tasks of low cognitive complexity, where AI acts as an assistant or controlled automator, and gradually build more autonomous systems (ceeim.es).
This approach allows for the development of transversal and informed adoption, with transparency, governance and a sustained learning curve.
Conclusion: AI to be more competitive safely
Manuel Romero's analysis highlights that AI has real potential to increase competitiveness, provided it is implemented with reliability, traceability and economic sense. It is not about adding complexity, but about improving processes where it really adds value.
For SMEs and tech freelancers, the message is clear: adopt AI progressively, measure its impact, guarantee transparency and move forward with human supervision and control mechanisms. If you want an expert team to accompany you on this path, CADT Software is ready to help you transform your processes with realistic, efficient and safe AI.
