From co-pilot to architect: how AI is reshaping the future of work
28/10/2025
“The challenge is no longer to test the technology, but to integrate it with criteria and strategy”. With this phrase, Raúl, AI Strategic Consultant at Pasiona, sums up one of the biggest gaps in the adoption of AI in Spain: many companies play with isolated pilots, but few manage to transform the way they work.
In this interview, Raúl talks about the most common mistakes, how to move from experiments to real impact, and why solutions like AIgents Manager and MIA are key to turn AI from a co-pilot to a process architect.
Q1: Many companies in Spain are already implementing AI, but not always successfully. What would you say are the most frequent mistakes you observe in this adoption?
The most common mistake is to think that it is enough to “put an internal ChatGPT” to get immediate results. Generative AI is very powerful, but without integration with corporate systems it remains a flashy demo. For example: an assistant that is not connected to ERP, CRM or internal documentation can only respond with general information and not with useful data for the business.
Another common failure is to underestimate data preparation and prompt design. Generative AI does not invent magic solutions: it needs reliable sources and a well-defined context to provide useful answers. Many companies fail because they upload poorly structured documents without metadata or quality criteria, and then blame the model.
And, of course, there is the human factor: generative AI changes the way we work, and if teams are not trained in how to interact with it, the tool is perceived as a hindrance rather than a co-pilot.
Another common failure is to underestimate data preparation and prompt design. Generative AI does not invent magic solutions: it needs reliable sources and a well-defined context to provide useful answers. Many companies fail because they upload poorly structured documents without metadata or quality criteria, and then blame the model.
And, of course, there is the human factor: generative AI changes the way we work, and if teams are not trained in how to interact with it, the tool is perceived as a hindrance rather than a co-pilot.
Q2: At Pasiona you have created tools such as AIgents Manager and MIA. What role can these solutions play in making AI not just a “hype” but a real strategic driver in organisations?
AIgents Manager and MIA are the step from fashion to strategy and will gradually grow to support the needs of the company.
- AIgents Manager allows the coordination of multiple generative agents within the organisation. One agent can be responsible for writing reports based on business data, another for answering frequent employee queries and another for preparing business proposals. With AIgents Manager they are all orchestrated, controlled and aligned with the company’s objectives in an easy way and without major technical requirements so that each area can manage its agents autonomously.
- MIA is the gateway to generative AI for any department. It enables HR to write job descriptions in seconds. HR to write job descriptions in seconds, marketing to create variations of campaigns tailored to different audiences or finance to prepare executive summaries of long reports. All this in a secure environment, connected to corporate systems and with traceability of use.
The role of these solutions is clear: to turn generative AI from a curiosity into a critical infrastructure for productivity and strategy.
Q3: The dominant narrative is about “co-pilots” assisting workers, but increasingly we hear that AI is starting to make decisions and prioritise. Are we now entering the era of “digital invisible managers”?
With generative AI we are starting to see this. Until now, copilots were assistants who helped to write, summarise or translate. But there are already applications that reorganise agendas, suggest which emails to attend to first or generate action plans from scattered documentation.
An example: a generative agent connected to Jira and Confluence can automatically reorganise the backlog of a development team, prioritise tasks and prepare a summary for the steering committee. Here we are not just talking about attendance, we are talking about invisible decisions that set the pace of work.
The challenge is to maintain human control: AI can propose and prioritise, but the final say must be with people. The interesting thing is that with generative, these proposals come in natural language, which makes them much easier to interpret and discuss in a team.
An example: a generative agent connected to Jira and Confluence can automatically reorganise the backlog of a development team, prioritise tasks and prepare a summary for the steering committee. Here we are not just talking about attendance, we are talking about invisible decisions that set the pace of work.
The challenge is to maintain human control: AI can propose and prioritise, but the final say must be with people. The interesting thing is that with generative, these proposals come in natural language, which makes them much easier to interpret and discuss in a team.
Q4: One of the biggest fears is that AI will end up deciding on productivity and even employment. Where is the red line and how does Pasiona ensure that AI is used in a fair and responsible way?
The red line is clear: generative AI should not be the judge in decisions that directly affect people. It can help prepare performance reports, provide feedback, or generate activity metrics, but it can never decide on promotions or dismissals.
At Pasiona, we ensure that any use of generative AI is supervised by a human and meets ethical and transparency criteria. This entails:
At Pasiona, we ensure that any use of generative AI is supervised by a human and meets ethical and transparency criteria. This entails:
- Let users know when a text, recommendation, or report has been written by a model.
- Systems should be auditable: If an agent generated a recommendation, there should be a record of what information was used.
- There should always be human review of critical processes.
Generative AI should be a tool for empowerment, not control. Its role is to free up time and provide clarity, not to become a judge of performance.
Q5: There is a lot of talk about cost savings, but rarely about the hidden costs of implementing AI without prior diagnosis. What are the real risks for companies that do not plan for such integration?
In generative AI, hidden costs tend to come from two fronts:
- Poor integration. If a chatbot is deployed without connecting it to the document repository or CRM, it remains a generic tool that provides no real value. This leads to abandonment of the tool and mistrust in future AI initiatives.
- Lack of governance. Without a security framework, AI can leak sensitive information or generate inconsistent responses. This has not only technical costs, but also reputational costs.
That is why we always recommend starting with a diagnosis: which sources of information will be used, which use cases are a priority, how the results will be monitored. Without such an analysis, the risk is to have an “expensive toy” instead of a strategic co-pilot.
Q6: For IT professionals involved in AI projects at Pasiona, what learning and growth opportunities arise that they might not find in other companies?
Working on generative AI projects at Pasiona means being at the forefront of the technological revolution.
Our teams learn to design complex prompts, to build autonomous agents that combine different models, to apply RAG (Retrieval Augmented Generation) techniques to connect LLMs with corporate knowledge bases and to deploy secure solutions in production.
Moreover, the value is in the integration. It is not just about using a pre-trained model, but connecting it with business systems (ERP, CRM, intranet) and making it useful for employees and customers. That hands-on experience, combining the latest in generative AI with real corporate environments, is a learning differentiator.
Our teams learn to design complex prompts, to build autonomous agents that combine different models, to apply RAG (Retrieval Augmented Generation) techniques to connect LLMs with corporate knowledge bases and to deploy secure solutions in production.
Moreover, the value is in the integration. It is not just about using a pre-trained model, but connecting it with business systems (ERP, CRM, intranet) and making it useful for employees and customers. That hands-on experience, combining the latest in generative AI with real corporate environments, is a learning differentiator.
Q7: Looking ahead, do you think AI will continue to be a supportive co-pilot or will it completely reshape the way we work and organise ourselves within companies?
Generative AI goes far beyond the co-pilot. Today we already see it creating presentations, reports, business proposals or marketing campaigns in a matter of seconds. It is support, yes, but at a speed and scale that changes the rules of the game.
In the medium term, we will see generative AI start to redesign entire processes. In one company, a generative agent may be responsible for dealing with suppliers, preparing preliminary contracts and coordinating meetings, while in another it may automate technical documentation and internal support.
This means that not only what we do changes, but also how we organise ourselves: fewer hierarchies, smaller and more agile teams, people dedicated to creativity, strategy and relationships, while AI takes care of the “heavy lifting” of content generation and coordination.
In short: generative AI starts as a co-pilot, but its destiny is to be the architect of new ways of working.
In the medium term, we will see generative AI start to redesign entire processes. In one company, a generative agent may be responsible for dealing with suppliers, preparing preliminary contracts and coordinating meetings, while in another it may automate technical documentation and internal support.
This means that not only what we do changes, but also how we organise ourselves: fewer hierarchies, smaller and more agile teams, people dedicated to creativity, strategy and relationships, while AI takes care of the “heavy lifting” of content generation and coordination.
In short: generative AI starts as a co-pilot, but its destiny is to be the architect of new ways of working.
Q8: What real difference does generative AI make to previous waves of automation?
The difference is radical. Classic automation focused on repetitive tasks with fixed rules: if “A” happens, do “B”. Predictive AI, on the other hand, analysed data to anticipate outcomes: predict customer churn, detect fraud or estimate future sales.
Generative AI breaks completely new ground: it doesn’t just execute or predict, it creates. It writes reports, generates images, synthesises conversations, develops code or even designs business proposals.
In other words, while previous technologies optimised what already existed, generative AI introduces the ability to innovate from scratch, democratising creativity. For example, a marketing department no longer relies on weeks to prepare a brief and several iterations with agencies: in minutes it can have ten versions of a campaign and decide which to scale.
Generative AI breaks completely new ground: it doesn’t just execute or predict, it creates. It writes reports, generates images, synthesises conversations, develops code or even designs business proposals.
In other words, while previous technologies optimised what already existed, generative AI introduces the ability to innovate from scratch, democratising creativity. For example, a marketing department no longer relies on weeks to prepare a brief and several iterations with agencies: in minutes it can have ten versions of a campaign and decide which to scale.
Q9: Generative AI is advancing at a rapid pace – how can a company avoid falling behind without improvising?
The trick is to balance speed and strategy. Going too slow means losing competitiveness; launching without a plan leads to failed projects.
Our recommendation is to adopt a “strategic pilots” approach: well-defined generative AI use cases, with rapid and measurable impact, but designed to scale. For example, an internal assistant summarising legal documentation. It’s a limited project, but one that demonstrates value, saves real time and builds confidence in teams.
From there, a maturity roadmap is built: connecting AI to more systems, extending use cases to other departments and creating a governance framework. In this way, the company moves at the pace of innovation, but with control and a long-term vision.
Our recommendation is to adopt a “strategic pilots” approach: well-defined generative AI use cases, with rapid and measurable impact, but designed to scale. For example, an internal assistant summarising legal documentation. It’s a limited project, but one that demonstrates value, saves real time and builds confidence in teams.
From there, a maturity roadmap is built: connecting AI to more systems, extending use cases to other departments and creating a governance framework. In this way, the company moves at the pace of innovation, but with control and a long-term vision.
Q10: And finally, what would be your recommendation for a company that wants to start implementing AI?
My recommendation is to start with a progressive and pragmatic approach, choosing areas where generative AI delivers immediate and scalable results. Not all companies have to start with the same thing, but it is worth designing a phased roadmap. One proposal would be:
- Office productivity. This is often the most natural entry point. With tools such as Microsoft 365 Copilot, Google Duet AI or Notion AI, any employee can save hours on writing emails, summarising meetings, making presentations or analysing spreadsheets. The ROI here is almost immediate: less time on repetitive tasks, more focus on the strategic.
- Business and decision-making. The next step is to integrate generative AI into business systems. For example, connecting an assistant to Dynamics 365, Salesforce or ERP to automatically generate personalised business proposals, financial summaries or executive reports. The key here is Retrieval Augmented Generation (RAG), which allows AI to access internal company data to provide relevant and reliable answers.
- Customer experience. This is a critical area where generative AI is already making a difference. Advanced chatbots that not only answer questions, but understand the customer’s context and write natural messages; personalised content generation for marketing campaigns; or even voice assistants that take orders in real time. Here we see deployments with Azure OpenAI Service or Dialogflow to deliver seamless and secure experiences.
- Internal knowledge and employee support. Many organisations have thousands of pages of documentation, manuals or procedures that few people read. A generative agent connected to this document base can answer any internal process question in seconds. This drastically reduces search time and encourages team autonomy. Microsoft Copilot Studio, for example, allows these agents to be built in a controlled and secure way.
- Innovation and creativity. Generative AI is also a catalyst for new ideas. From creating product prototypes in a matter of hours to writing business plans or strategic proposals in record time. Tools like ChatGPT Enterprise or Azure AI Studio allow innovation teams to iterate faster and test concepts that used to take weeks.
- Operations and internal efficiency. Another field with high impact is that of process automation with generative AI. For example, agents that automatically write meeting minutes, translate contracts in real time or generate technical documentation from specifications. Here AI not only speeds up, but also standardises and reduces human errors.
- Compliance and governance. Finally, it is important not to forget the security and compliance framework. Many platforms (Microsoft, AWS, Google) already include layers of traceability, data control and auditing that allow companies to ensure responsible use of AI. The recommendation is always to deploy with governance from the outset to avoid future risks.
In short: start with productivity to build trust, then scale to business and customer, and finally extend to innovation and operations. The key is to combine visible use cases – that excite teams – with a solid architecture that allows for controlled growth.
Generative AI is not an end in itself, but a tool to redesign the way companies operate and make decisions. At Pasiona, with AI Assessment, we accompany organisations to move from isolated testing to a sustainable model of adoption, with a focus on strategy, scalability and tangible value.
AI co-pilot, Aigents Manager, generative artificial intelligence, MIA, pasiona, Spanish companies
Go backYou may also like
AI in business: key lessons from the report “Superagency in the workplace”
25/11/2025
McKinsey report reveals the contradictions in AI adoption in the enterprise. We look at the most relevant data and what it means for leaders and employees.
Who fires whom? The day an AI agent says you’re redundant in the company
18/11/2025
AI is no longer just analysing data: it is starting to suggest reorganisations. In Pasiona we explain how to govern it so that it complements, not replaces, human decisions.
The AI that rules more than your boss
14/10/2025
AI is no longer just a companion: it is starting to decide. At Pasiona we tell you how to govern the new digital managers before they take over.