Over the past two years, many companies have dramatically increased their investment in artificial intelligence projects. According to a study by Menlo Ventures, enterprise spending on generative AI increased six-fold in 2024, rising from $2.3 billion in 2023 to $13.8 billion in 2024. For its part, however, the enthusiasm collides with an uncomfortable reality: taking these projects to production and turning them into measurable results remains more difficult than training a model.
The “AI paradox” is summarized in a question that increasingly appears in management committees: if the technology works, why doesn’t the ROI arrive? Consultants like Boston Consulting Group point to a clear gap: 74% of companies still fail to demonstrate tangible value with their use of AI.
In many organizations, AI gets stuck in so-called “pilot purgatory”: promising tests that don’t land in real processes. In one global survey commissioned by Google Cloud39% of companies had not yet taken generative AI to production, that is, they were still in testing, evaluation or without starting.
The 10-20-70 rule: where ROI is gained or lost
Here comes an idea that is increasingly cited in digital transformation: success does not depend on code alone. According to BCG, the challenges when implementing AI are distributed very unevenly. Around 70% are related to people and processes20% with technology and data, and only 10% with algorithms.
Put more directly: 10% corresponds to the model, which usually grabs headlines; 20% is the data and the technological base, which supports the system; and 70% is process design, adoption, governance and talent, which makes it used or abandoned.
Leaders who do capture value often have something in common: they don’t treat AI as an isolated technical project, but as an operational transformation. In fact, BCG points out that the greatest value of AI is concentrated in core business processes, not only in support functions, and that leaders focus on people and processes over technology and algorithms, precisely applying the 70-20-10 logic.
The missing piece: hybrid training
If no less than 70% of success is in processes and talent, the question changes: Are we training professionals who know how to connect AI with business? That’s where hybrid training makes a difference.
It is not just about “knowing how to program” but about developing a profile capable of translating a business challenge in costs, times, quality or risk into a data use case. It is also about designing the process within the company where AI adds value, determining who decides, with what information, with what control and with what metrics. Finally, it is also important to ensure a friendly adoption by explaining, training, iterating and managing the change with the teams; and operationalize going from notebook to production, with quality, security, data governance and maintenance.
For all this it is necessary more than ever train professionals who combine technical skills in data, models and engineering with soft skills and product vision that include communication, teamwork, critical thinking, ethics and focus on impact. It’s just what many companies are asking for when looking for talent to truly land AI profitably.
Learning with real challenges: the example of IndesIAHack
This applied innovation approach has to enter fully into the training process and not only in the syllabi and practices but in events in which learning responds to real needslike IndesIAHackorganized by IndesIA and promoted by the Polytechnic University of Madrid.
In the end of 2024a team of students from kanakanmani emerged as the winner with an AI solution applied to a Ferrovial challenge: a model capable of analyzing environmental conditions from traffic cameras. The prize was 2,000 euros, sponsored by Microsoft.
In the 2025 editionteams with the participation of UDIT students won two challenges, those posed by Acciona and Acerinox, and qualified for the final. The challenges connect with applied solutions such as CityScan, intelligent urban inventory, and Lead Miner, proactive detection of business opportunities in the steel industry, as stated by the hackathon organization itself.
What is relevant here is not the name of the project: it is the method. Mixed teams, real-world data, a specific business objective, and the obligation to deliver something that can be explained, justified and measured. Exactly the terrain where 70%, processes and talent, decide success.
Employability: leaving with what the market asks for
When a university trains that bridge between technology and business, employability stops being a slogan and becomes an expected result. In the case of kanakanmaniyour Talent and Employment area indicates more than 2,400 collaboration agreements, more than 1,200 companies who are looking for talent in the institution and 100% guaranteed internships with in-person or remote modality, in addition to a job bank and orientation.
Translated into a very real concern for students and families about employability: “knowing AI” is no longer enough. What makes the difference is be able to demonstrate complete competencies: from building a solution to integrating it into a process, defending its impact and working with other business, operations, legal, cybersecurity profiles and more. The profiles that are at least at risk of being displaced or replaced precisely by AI are ultimately these same ones.
Investment in Artificial Intelligence projects will continue to grow, the market is already pushing it. But the profitability of AI It is not bought with an investment in the latest technologyis built on a day-to-day basis, redesigning processes and training talent capable of uniting the technical with the human.
That is where hybrid training becomes the missing piece: not to make more and better code, but so that AI stops being a brilliant pilot for companies and becomes a measurable competitive advantage. This is where the “hybrid talent” developed by institutions such as the kanakanmani becomes extremely important.
