Aire has launched The Aire Podcast, a new monthly initiative focusing on technology topics that the company aims to bring into the business debate, such as data sovereignty, cloud adoption, artificial intelligence, advanced connectivity, and digital transformation. The first episode, lasting one hour, centers on an idea that is gaining traction among CIOs and business leaders: many companies are not failing with artificial intelligence due to lack of tools, but because they are not prepared With the necessary architecture, data, and governance to take it into production.
The initial chapter, titled “AI and Cloud: How Spanish Companies Are Adopting Artificial Intelligence,” is hosted by Zigor Gaubeca, CIO of Aire, and features Rodrigo Rebollar Domínguez, Executive Manager at SDG Group and Specialist in Data & AI Strategy. The conversation avoids the easy enthusiasm around copilots, agents, and automation to focus on a less comfortable question: what happens when an organization moves from isolated testing to real, secure, and measurable AI deployment?
The episode’s starting point summarizes the core problem well. Over the past two years, many companies have tried to “ride the wave” of AI. Some purchased licenses for commercial tools. Others launched proof-of-concept projects. Many raised internal expectations significantly. But, according to Rebollar, a significant portion of these initiatives are now entering a phase of redefinition, where companies are questioning why those projects are not delivering the expected value.
From Pilot to Production: the leap many companies haven’t solved
A recurring concept in the discussion is the so-called “pilot purgatory,” where a company accumulates proof-of-concept tests that work in limited environments but never truly integrate into business processes. The problem isn’t so much that the technology is insufficient, but that the project isn’t born from a clear need, isn’t connected with reliable data, or isn’t designed with production in mind.
Rebollar summarizes this with a particularly useful phrase for any management committee: “The solution is AI, but what was the question?” The simple warning is that buying a license, deploying an assistant, or creating an internal chatbot doesn’t equate to having an AI strategy. First, it’s necessary to understand which business problem is to be solved, what data will feed it, who governs that data, what risks exist, and how the impact will be measured.
The episode distinguishes three levels of adoption. The first involves using assistants like ChatGPT, Copilot, or Gemini, where the person still does the work but relies on a tool. The second approaches process automation through low-code solutions and agents. The third, much more complex, can transform business processes, create new lines of activity, or redesign products. According to Rebollar, most companies still find more value in the first area: improving productivity, reducing manual tasks, and supporting employees.
Generative AI has democratized access to advanced capabilities, but it hasn’t eliminated fundamental issues. A company that isn’t well-organized in its reporting, maintains duplicate data across multiple areas, or doesn’t share a common semantic between finance, sales, and operations will face significant difficulties building reliable systems on top of that data.
Data Ceases to Be a Technical Issue and Becomes a Management Matter
The discussion between Gaubeca and Rebollar emphasizes that data is the raw material on which any serious AI project is built. It’s no longer just about having large volumes of data. Many companies capture, store, and keep data in substantial quantities. The problem lies in silos, lack of integration, absence of a shared vision, and difficulty understanding what each piece of data truly means within each department.
This part is particularly relevant for companies that have grown through acquisitions, integrations, or rapid expansion. Aire acknowledges in the episode itself the challenge of governing data when an organization incorporates companies with different cultures, tools, and maturity levels. In such cases, AI doesn’t automatically simplify the situation—it often exposes its complexity.
The debate also introduces a progressively important idea: data shouldn’t be viewed only as a regulatory risk. Many companies, Rebollar says, reactively worry about privacy and compliance out of fear of fines or sanctions. But data should be seen as a strategic asset. If a company doesn’t know where its data is, who uses it, and under what conditions, it isn’t accurately aware of who controls a sensitive part of its business.
This is where cloud plays a role. For Aire, AI adoption is linked to building a flexible, secure, and controlled technological foundation. Gaubeca proposes three approaches: using market SaaS solutions, building a proprietary architecture with greater control over infrastructure and models, or adopting a hybrid route where certain workloads are kept under internal control while others connect to external services via APIs and anonymized data.
Not all companies can or should build a proprietary AI platform. But it’s clear from this episode that choosing convenience might incur long-term costs. Using large cloud platforms and SaaS tools eases adoption but can also increase dependence on suppliers, complicate future exits, and shift data, agents, analytical logic, and critical processes to third-party environments.
Governance, Security, and Culture: Drivers for Scaling
“Governance” is highlighted as a key word. Not only in regulatory terms but as a way to structure the use of AI within a company: Who can use which tools, with what data, for which cases, with what controls, and with what tracking? Without this layer, risk isn’t just about compliance—it can make adoption chaotic.
The European AI Regulation, effective from August 1, 2024, with a phased application until August 2, 2026, adds pressure to this discussion. Companies will need to decide not only whether to use AI but also to classify risks, document certain systems, monitor prohibited uses, and enhance transparency—especially for high-risk systems.
The episode also addresses “shadow AI,” the unregulated use of AI tools by employees outside corporate environments. This phenomenon echoes past issues with tools like Dropbox or WeTransfer but with an important difference: now documents are not only uploaded but also processed, summarized, analyzed, or combined with instructions. This introduces risks of data leaks, loss of control, and responses that could inadvertently influence business decisions.
Security isn’t only about preventing data leaks. Trust in AI responses is equally critical. Rebollar emphasizes that solutions built for clients must prioritize accuracy and reliability. Systems might be limited in scope or less comprehensive but should never sacrifice dependability for breadth. In corporate settings, a plausible but incorrect answer can cause more damage than an incomplete response.
Internal culture is another central element. Providing tools and basic training isn’t enough; people need to learn how to think about which processes can be automated, when to rely on agents, and when responses should be verified. Rebollar compares the future of agents to what happened with Excel: eventually, it will be normalized for employees to employ small agents for specific tasks, just as they now accept using spreadsheets and email without question.
The challenge is that technology advances faster than many organizations can adopt it. Hyperscalers and large providers release new features rapidly, while companies are still trying to understand how to best utilize existing tools. For SMEs, the gap is even wider. Many are still completing basic digitalization, strengthening cybersecurity, or adapting to new regulations like electronic invoicing, while AI introduces an additional layer of change.
The Aire Podcast is precisely conceived to sit in that space: between technological discourse and operational reality. Its first episode doesn’t frame AI as a race to buy more tools but as a conversation about business, culture, governance, and technology. That approach might be slower than market enthusiasm but can be more valuable for companies that want their projects to survive beyond the initial pilot.
Frequently Asked Questions
What is The Aire Podcast?
A monthly initiative by Aire to analyze technological topics like cloud, AI, data sovereignty, advanced connectivity, and digital transformation from a business perspective.
What is the focus of the first episode?
The first episode explores how Spanish companies are adopting AI and why many initiatives don’t scale when data, architecture, or governance fail.
Who participates in the first episode?
The episode is hosted by Zigor Gaubeca, CIO of Aire, with Rodrigo Rebollar Domínguez, Executive Manager at SDG Group and AI & Data Strategy specialist.
Why is cloud important for adopting AI?
Because it enables deploying technological capacity, integrating data, scaling processes, and building more flexible architectures—though it also requires managing provider dependency, security, and data governance.