After the initial phase of AI adoption, it is time to demonstrate what it is for. After a first stage marked by experimentation with tools, proofs of concept and pilot projects, companies are beginning to enter a more demanding phase: AI has to be translated into concrete results for the business.
Saving time, reducing costs, automating tasks, improving processes or facilitating decision making is beginning to outweigh the simple fact of incorporating the latest technology available. This is what they are detecting from Zenital, a boutique consultancy specialized in data and artificial intelligence, based on their work with companies and the training given to more than 200 professionals in AI during the last year.
The change does not mean that companies have stopped experimenting, but rather that the questions are different. Faced with ‘what can we do with AI?’, questions such as ‘what process do we want to improve?’, ‘how much time can we save?’, ‘what does it cost?’, ‘what data do we need?’ are beginning to gain ground. or ‘how are we going to measure if it really works?’. Added to these is an increasingly relevant challenge: how to guarantee that the adoption of AI is real, sustainable and transversal to the entire organization, and is not limited to isolated initiatives or certain teams.
“In recent years there has been enormous curiosity about artificial intelligence and that has been positive because it has allowed companies and professionals to lose their fear of experimenting. But we are entering a second stage. It is no longer enough to say that we use AI. We must demonstrate where it adds value and what result we achieve with it,” explains Manel Vericat, Data & AI Training Specialist at zenital.
When AI moves from test to process
The difference between experimenting with a tool and making it part of a business process lies precisely in the ability to measure its impact. A recent project developed by Zenital to automate the collection and prioritization of a team’s training needs allows us to put figures to this change. This task initially required about 240 hours of work and, through the application of AI, it is estimated that it can be reduced to 80 hours, which would free up around 160 hours of work.
Beyond economic savings, the example reflects one of the main changes that Zenital observes in organizations: the need to link the adoption of AI to indicators that allow objective evaluation of whether the project is adding value. For the consulting firm, the real leap occurs when a company is able to identify a task or process, establish how much it currently costs in time, resources or money and compare that starting point with the result obtained after introducing the technology.
“A proof of concept (PoC) can be spectacular and yet not solve any real problem for the company, or provide less value than what the implementation and consumption of tokens would cost. The challenge is to identify where AI applied to the business saves costs without reducing reliability or control. Sometimes it will be to automate a task; other times, so that a person can make a decision with more information or dedicate their time to activities of greater value. And there will be situations in which the correct response will be not to use it,” says Manel. Vericat.
Knowledge, organization and data
Achieving effective AI adoption doesn’t just depend on having the right tools. Based on its experience with organizations and professionals, zenital identifies four recurring obstacles: lack of knowledge of the available tools, a limited understanding of the technology itself, the rigidity of organizations and the speed at which AI is evolving.
The lack of knowledge can also cause the opposite effect to that expected: overestimating the capabilities of artificial intelligence and ignoring its limitations and risks. The speed with which new models, functionalities and ways of working appear at the same time makes it difficult for companies and professionals to consolidate knowledge before the scenario changes again.
Added to these challenges is the quality of the information. Data quality and governance become especially important as companies move from individual uses of generative AI tools to applications connected to corporate information and processes.
An AI can be as sophisticated as technology allows, but if it works on incomplete, duplicate, outdated or poorly structured information, its results will inevitably be conditioned by the quality of that raw material. Zenital summarizes it with a simple idea: applying artificial intelligence on poor quality data does not solve the problem, but rather amplifies the chaos.
For this reason, the consulting firm highlights the importance of auditing the data before addressing the implementation of artificial intelligence solutions, with the aim of evaluating their quality, detecting possible inconsistencies and determining whether the available information is really prepared to feed these systems. A need that, according to Zenital, is gaining weight among organizations: the company is recently seeing an increase in demand for this type of audits as a prior step to incorporating AI into its processes.
“The appearance or incorporation of AI in business processes is making many organizations look at their data with different eyes and begin to realize that perhaps all that dispersed information should begin to be cleaned. It is a natural process: I try to solve something with AI, it doesn’t work because the data is not right, and then I put the focus on the data,” explains the Zenital specialist.
From learning to make prompts to transforming processes
Change is also coming to business training. During the last year, Zenital has trained more than 200 professionals from around 15 companies in artificial intelligence, mainly in the areas of Sales, Operations and Administration. This activity is allowing the consultancy to observe first-hand how the needs of organizations evolve. If the first approaches were very focused on knowing the tools and learning to interact with them, interest is now beginning to shift towards an issue much more linked to the business: how to use AI to optimize processes.
This involves identifying which tasks are worth automating, how to work with corporate information, what limits should be established or how to ensure that the adoption of AI is not restricted to a small group of employees especially interested in technology. “Training is beginning to change because companies no longer just want their teams to know how to use a tool. They want them to understand how to incorporate it into their work with criteria. The objective should not be to do with AI the same thing that we did before a little faster, but to review processes and ask ourselves if we can work in a better way,” says Manel Vericat.
From generative AI to agents
This change comes as technology continues to advance. The evolution of generative AI models and the development of systems capable of executing increasingly complex tasks are opening up new possibilities for organizations, but they are also raising demands regarding security, integration, governance and control.
Within this scenario, zenital is reinforcing its specialization in Claude, the artificial intelligence developed by Anthropic. Currently, ten professionals from the company have the Claude Certified Architect Foundations certification and the consultancy works on different use cases with this technology, while advancing in the process to become an Anthropic’s Claude Select Partner.
During the last year, Zenital has trained more than 200 professionals from around 15 companies in artificial intelligence
For Zenital, however, the challenge goes beyond incorporating new tools. The next phase of AI adoption will be determined by the ability of organizations to identify where it adds value, integrate it into their processes, train their teams and establish a governance framework that allows its use to be extended safely and effectively.
In 2027, AI will have to justify the investment
Looking ahead to 2027, Zenital foresees that business artificial intelligence will enter a stage of greater maturity, in which proofs of concept will increasingly coexist with demands for return, scalability, security and control. The adoption of AI will thus no longer be measured solely by the number of tools used or projects launched and will begin to be measured by its impact on specific indicators such as hours saved, reduced costs, automated processes, response times, productivity or improved analysis and decision-making capacity.
In this context, the adoption of AI will become increasingly linked to the ability of companies to justify the investments made and demonstrate that the technology generates a tangible impact on the business. The question will no longer be only how much AI is used, but how much value it contributes and with what guarantees.
“The hype has forced the entry of artificial intelligence into the agenda of practically all organizations. Now comes the difficult and, at the same time, the most interesting part. We have to separate what is technically possible from what really makes sense for each business. The companies that do this exercise well will be the ones that manage to turn AI into an advantage and not simply another tool. To achieve this, at least two things are necessary: training and governance,” concludes Manel Vericat.
