This year, artificial intelligence has ceased to be a promise and has become a real working tool. Yes, I know this sentence could appear in any year-end summary, but in our case, it is exactly what has happened. And I (the one writing this post is me, Anna) found it rather amusing to see how at DeMomentSomTres we have been incorporating it into our daily routine almost without ceremony: not as an embellishment, but as another tool to work better.
The difference, however, is not merely having AI. The difference lies in integrating it with discernment and method into real processes and teams—those that involve emails, urgencies, people, and decisions. And maintaining a central idea: improving results without losing the human perspective, which remains what makes things work.
From Discussion to Implementation: AI as a Tool for Transformation
AI is not a demiurge: it does not create from nothing. However, it does accelerate when you have discernment, data, and a clear process. If we had to summarize 2025 in one idea, it would be this: AI has transitioned from being a ‘topic to explore’ to a daily lever for improvement. And this has not been about having the latest tool, but about understanding what needed to be solved in each organization and how to make the solution fit into its daily operations.
This evolution has also been noted in how clients approach us. Less and less do we hear ‘we want to do an AI project,’ and increasingly the correct question arises: ‘Where does it make sense to apply it here, and how do we do it without disrupting the organization?’
AI in over 200 projects at DMS3
Throughout the year, AI has been involved in over 200 projects in one way or another. At times, it has been visible and central; in others, it has been a discreet layer that accelerates tasks, improves quality, and helps in making decisions with more information.
The common denominator is the same: we do not talk about AI ‘just because it’s trending,’ but about AI with a clear function and realistic integration.
From Sant Jordi to Critical Processes: Concrete Examples of What We Have Built
When discussing AI, it is easy to remain in generalities. That is why we prefer to explain it with real cases, very different from each other, but with a shared pattern: a specific challenge, a clear process, and an adoptable solution.
A Haiku Tool for Sant Jordi
For Sant Jordi, we created a tool capable of generating haikus with artificial intelligence from a simple theme. It is an apparently simple, but very illustrative example: a minimal instruction can become a coherent, controlled, and customizable creative piece.
And this same principle—clearly defining the intention and delimiting the result to be useful—is what is then applied in more demanding environments.
AI Agents for Businesses
We have also set up agents for businesses designed to support specific teams and processes: assistants that help retrieve information, guide decisions, prepare responses, or execute repetitive steps with a defined logic.
Not as a ‘technological toy,’ but as a support layer that reduces friction, increases consistency, and frees up time for higher-value work.
Customer Service Chats with Quality Standards
We have deployed customer service chats aimed at improving responsiveness and availability. However, there is an essential condition: a chat only provides value if it is connected to reliable content, if it knows when to escalate, and if it maintains quality standards and brand tone.
The goal is not to automate for automation’s sake, but to ensure a better and more consistent experience.
Email Classification When Volume Is a Real Problem
In organizations with a high volume of incoming information, we have designed processes to classify large volumes of emails by importance and topic.
This is one of the clearest cases where AI can provide immediate gain: reducing noise, improving prioritization, accelerating internal referrals, and minimizing the risk of a critical request being buried. But for it to work, criteria must be defined (what does ‘urgent’ mean here?), circuits (who receives what?), and validations (how do we control errors?).
Training, Certifications, and Discernment: It’s Not About Collecting Tools
This year, we have also strengthened our knowledge base with training and certifications, including reference training such as that from MIT, and we have expanded our discernment by testing and comparing over 20 different solutions and approaches.
Not for collecting tools, but to be able to rigorously respond to a reality: there isn’t a single AI that works for everyone. The same technology can work very well in one environment and be a source of friction in another.
The Key Point: AI at the Service of Humans and Teams
The most important aspect, however, is the mindset. AI does not replace people; it transforms how they work. And this demands respect for the context: how the team is organized, which processes are critical, what information circulates, what risks cannot be assumed, which decisions require professional judgment, and which tasks are repetitive and automatable.
At DMS3, we work with AI as a tool at the service of humans and specialized teams. This means:
- understanding the real process (not what the procedure states, but what actually happens)
- defining a viable fit (data, budget, maturity, and culture)
- ensuring adoption (criteria, training, guidelines, and quality measures)
If a tool is not used, it is not innovation; it is an expense.
Closing the Year with a Clear Idea
We close 2025 with a very operational conviction: AI is a primary tool, yes, but the difference lies in how you integrate it into an organization without losing the human perspective.
2026 will not be about ‘doing AI,’ but about making it useful.
