The key role of artificial intelligence in the evolution of APIs and digital integration
Artificial intelligence plays a key role. When we talk about evolution, we are talking about maintaining and developing new tools that offer greater capabilities. In companies such as Telefónica, which has one of the largest API catalogues in the country, the use of AI to optimise this evolution is crucial to ensuring we do not fall behind.
When developing a microservice, or indeed any other software at Telefónica, generative AI acts as an unprecedented accelerator, drastically reducing time-to-market.
Its true value goes beyond simply speeding up development, as it enables us to build systems whilst ensuring compliance with the company’s various guidelines and best practices – which is one of the greatest challenges when working with such an extensive catalogue, on which different integrators operate.
Growing quickly is essential, but the biggest challenge is being able to do so in an organised and easily maintainable way in the long term.
If AI enables us to do this in a much more automated and rapid manner, that is where we will be making the biggest leap forward.
How generative AI and intelligent agents are transforming systems integration
To understand the difference, we need to distinguish between two levels: AI as a tool to assist and accelerate, and AI agents that provide autonomous support.
‘Traditional’ AI, such as ChatGPT, Copilot or similar tools, acts as an accelerator that helps us in our work, mainly by explaining things or answering questions.
When integrating with systems we’re unfamiliar with, one of the biggest bottlenecks is documentation. Using AI allows us to turn technical documentation into assistants: “What fields are required for a call to the x API?” or “In what format is the data returned?”.
This makes documentation infinitely more accessible, speeding up our understanding of the systems and, consequently, the integration process.
In this vein, one of the most interesting practical uses is system mocking:
Instead of relying on limited and static test data, generative AI can be used to simulate the behaviour and responses of a service that already exists or has not yet even been developed, with the aim of carrying out integration tests in environments very similar to real-world ones, without having to wait for the infrastructure to be fully deployed.
Furthermore, the use of agents represents the next step beyond the AI we are all familiar with.
The revolution lies in the ability to give them control over our working environment, moving from an assistant to an autonomous system.
Instead of simply answering queries, we use them to execute entire workflows: they can use the terminal, install dependencies, run scripts, read error messages and resolve them independently.
Furthermore, they enable the creation of role-specific specialists:
All of this greatly accelerates digital integration and forces organisations to adapt so as not to be left behind.
APIs act as connectors for interacting with systems, retrieving data or implementing changes.
The key lies in the evolution from mechanical automation towards a model where AI interprets context, reasons and makes decisions in real time.
This enables us to transform traditional binary workflows (“yes or no”) into intelligent processes that analyse:
- whether the data makes sense
- whether the system is functioning correctly
- whether there may be faults or vulnerabilities
This paradigm shift improves efficiency in virtually any field.
How AI automates processes and improves operational efficiency at Telefónica
I recently had the chance to try out one of the most interesting use cases: integrating an AI agent into IT support at Telefónica.
There are many automated processes involved in onboarding new staff, but others require manual intervention.
In my case, I opened a ticket to access a portal and an AI responded directly, informing me that my user account had already been created.
Solutions like this enable the automation of recurring tasks, freeing up the support team and allowing them to focus their efforts on higher-value tasks.
It is a clear example of how to integrate AI into day-to-day operations.
Security, governance and data management challenges in AI systems
The use of AI speeds up processes, but it also introduces new vulnerabilities and increases the attack surface.
One of the main challenges is that AI generates responses based on the time at which it was trained. Best practices evolve and new vulnerabilities may emerge, so the generated code cannot go directly into production without validation.
There are also risks such as:
- Prompt injection to obtain sensitive data
- Probabilistic models that can fail or ‘hallucinate’: (To put it another way, the risk is that they will hallucinate because they are probabilistic models, not that the risk lies in their being probabilistic, as all models are probabilistic.) Telefónica’s systems require deterministic results (2+2 must always equal 4), but AI models are probabilistic and can hallucinate and make mistakes. ‘Guardrails’ are needed to ensure that responses always meet predictable standards.
- The need for deterministic responses in critical systems
It is therefore essential to implement guardrails to ensure reliable results.
In the case of agents, if they have too much access, they can cause serious errors such as data deletion.
They must therefore operate under the principle of least privilege. Another key aspect is the storage of data and models.
Many public solutions store information in the cloud and can learn from interactions.
This poses risks if confidential information is leaked.
That is why companies such as Telefónica use private infrastructure and strict agreements to protect data.
The evolution of the API analyst role towards strategic roles in artificial intelligence
In my experience, this role has evolved rapidly from a function centred on development and integration towards a cross-functional role, focused on innovation and with a strategic vision.
It is no longer enough simply to design interfaces. The focus is now on intelligent and autonomous systems, which requires new capabilities.
To achieve this, the following are key:
- a solid foundation in AI to integrate models
- knowledge of DevOps to deploy solutions
- designing CI/CD pipelines that automate the entire lifecycle
AI does not eliminate the need for technical roles; rather, it reinforces it.
It is now necessary to audit, validate and understand systems from a more strategic perspective.
Telefónica: leading the way in access to digital technology through innovation, trust and a European vision Telefónica is taking on the challenge of becoming the best gateway for citizens to access digital technologies, driving the company to become more innovative, competitive and customer-focused. This commitment is underpinned by European leadership, a contribution to technological sovereignty, the development of advanced services and the construction of the best network for accessing the most innovative technology. All of this with a clear objective: to offer more and better services, attract the best talent and build trust, recognising that quality is measured by the actual customer experience and the positive impact on society.
