ince its founding in 2008, Decide has embedded artificial intelligence into its technological DNA.
“We’re not a company that jumped on the AI bandwagon,” says Ángel Luengo, Head of AI at decide. “For over a decade, we’ve been helping companies make complex decisions using models that learn from data and integrate seamlessly into their daily operations.”
Unlike the often “hyped” narrative around AI, Luengo stresses that “it’s not magic” — but rather a technology that, when well designed, can transform data into complex patterns that enable efficient, measurable decisions aligned with business goals.
“Our approach has always been pragmatic. If it doesn’t solve a problem, it’s not a solution. The key is applying AI where it delivers real value.”
Agentic AI
Ángel Luengo advocates for drawing a clear line between their approach and traditional workflows that rely on language models.
“There’s a structural difference between a system that follows a predefined path and one that decides how to act in order to reach a goal,” he explains.
According to reference frameworks from organizations like Anthropic and OpenAI, agents begin where rigid orchestration ends.
They dynamically choose which tools to use, in what order, and adapt their steps based on the outcomes they observe. They don’t just respond they reason, act, observe, and readjust their behavior.
It’s a paradigm shift: it’s no longer about executing step-by-step instructions, but about building systems capable of making their own decisions based on a defined goal.
For Ángel, this has direct business implications: “The more agentic a system is, the more flexible and autonomous it becomes but also harder to audit, maintain, and govern. That’s why it’s essential to design with intent, and avoid applying agents where they don’t belong. You should never delegate what you can automate,” Luengo explains.
This philosophy translates into a layered architecture that Decide applies systematically.
A first layer of semantic validation using lightweight models and deterministic processes to ensure contextual coherence and access control;
a second layer of dynamic orchestration based on embeddings and adaptive planning;
and finally, a third layer that incorporates agents designed to solve specific tasks, integrated into a traceable and auditable environment.
Designing in layers allows us to tailor the level of intelligence and autonomy to each use case, without losing control or adding unnecessary complexity,” explains Ángel.
This modular approach has become a key competitive advantage, raising the success rate of pilots above 95%.
Decide has deployed operational solutions across diverse sectors such as energy, banking, insurance, and retail — always applying AI where it truly delivers value.
In the energy sector, for example, systems have been developed to dynamically adjust prices at charging points based on demand and market conditions.
In banking, intelligent chatbots are used to authorize transactions in real time while complying with the most stringent regulations.
In insurance, language models analyze unstructured documentation to extract key information, powering business processes and automating back-office tasks.
And in retail, intelligent systems are used to forecast demand, optimize stock replenishment, and plan staffing more efficiently.
“Over 95% of our developments make it to production because, from the start, we combine AI, architecture, and business in a coherent and applied way.”
“The key is to design every project from the ground up — with a clear purpose, a scalable architecture, and a strong business vision,” emphasizes Ángel.
What began as an experimental exercise with language models has evolved into high-impact operational services, proving that a structured, adaptable, and goal-oriented architecture is the most effective way to turn an AI pilot into a real production solution that generates tangible value.