Agentic AI Ben Jarvis David Small

Everyone is talking about AI, with the conversation turning towards “agentic AI” in recent months – but few business leaders are clear on how this changes things and what to do next.

 

As organisations move beyond pilot projects and chatbots, agentic AI is the next step towards the frontier firm, where “AI agents” work alongside humans to accelerate productivity.

 

But is agentic AI really new, and what do you need to consider? To unpack the concept, we spoke to two CTOs from Telefónica Tech, provider of digital transformation services across the UK: Ben Jarvis, CTO of Data & AI, and David Small, CTO of AI Business Solutions.

 

In this article, Ben and David discuss the what, the why and the how of agentic AI, answering your questions to lay a clear path forward for your business.

What’s the difference between agentic AI and other types of AI?

Q1. Let’s start simple: when most people say “AI” today, what are they usually referring to?

David: Many people still think of AI as chatbots like ChatGPT. They see it as asking a question and getting a response back – simple task‑and‑reply interactions. The hype has centred around these tools, so in my experience that’s what the general public tend to picture when they say AI. The companies that I see leading the way today are moving past this and they are redesigning work around agents.

 

Ben: People bundle everything together – generative AI or chatbots, large language models, machine learning, even basic analytics. But importantly, AI on its own doesn’t change anything. The real value comes from how organisations re‑engineer processes around these tools, which is what we need to reframe mindsets towards.

Data professionals collaborating on analytics solutions using Microsoft Fabric

Q2. What do we mean by “agentic AI”?

Ben: Agentic AI isn’t necessarily new, but it is a certain way to use AI that is gaining traction. Agentic AI is simply models combined with tools that let them act inside systems. It’s the integration and orchestration that make it feel more autonomous, rather than it being a completely new concept.

 

David: Expanding on that, agentic AI completes multi‑step workflows end‑to‑end. At least, that is what the term “agentic AI” refers to. Instead of gathering and generating information like a chatbot, an AI agent takes an action like sending something, triggering a process, or so on. It replaces multiple mundane steps rather than conducting a single task that a human still has to action.

What do AI agents actually do?

Q3. What does “agency” actually mean in this context - what can an agent do that a model or chatbot can’t?

David: A chatbot gives you information, but you still do the work.  An agent, on the other hand, can take a goal and carry a workflow through end-to-end. It can find and interpret data, decide the next steps, and take actions inside systems like booking time off, for example, or sending communications.  In practice, that means removing multiple manual steps rather than just supporting a single interaction. This doesn’t mean unchecked autonomy, though – humans still set the boundaries and define where intervention is required.

 

Ben: Agency is about giving AI tools it can use – the ability to gather data, act in systems, update records, book things, or trigger follow‑on tasks. It brings together intelligence and action instead of leaving everything for humans to finish off. Echoing what David says about removing human intervention, this is why governance and ethics are so important to implement when adopting AI across your organisation.

Why should business leaders pay attention to agentic AI?

Q4. From a business perspective, what are some examples that agentic AI can be used for?

Ben: Typically, businesses want three outcomes: make money, save money, or stay out of trouble! Agentic AI helps with all three. It creates new revenue opportunities, removes manual effort, speeds up processes, and adds auditability. This is true especially in areas like policing where decisions need to be consistent and traceable.

 

For example, we’ve developed a solution to support Clare’s Law, formally known as the Domestic Violence Disclosure Scheme, which allows police to disclose information about an individual’s history of domestic abuse where doing so is necessary to protect someone from harm. The solution enables officers to gather the required information up to three times faster, significantly improving response times when every moment counts.

 

David: Exactly that. Using AI agents frees people from repetitive tasks so they can focus on higher‑value work. Human‑in‑the‑loop checks are still there, but agents massively reduce time spent on the manual steps. It’s about capacity release and helping teams achieve more without extra budget. Some other interesting use cases I’ve seen recently include a Medicines Supply Tool agent in a healthcare setting, an SOP Navigator, and several HR related agents such as for onboarding.

Agentic AI Governance

Q5. “Autonomous AI” can sound uncomfortable. What concerns do leaders usually raise first?

David: Naturally, the first worry is always job displacement. People fear agents will replace them and their colleagues, but the aim should be to automate around the humans. Microsoft talks about this as the ‘journey to becoming a Frontier Firm‘ – humans are always in control. There is also the governance and visibility concern – how to audit actions, track what the agent has done, and ensure it isn’t making unmonitored decisions.

 

Ben: There’s also the ethical side. Leaders question whether autonomous AI is appropriate for decisions that affect people. They need traceability – exactly what data went in, what came out, and why. Using policing again as an example, cases like police forces being burned by hallucinations come up often, some of which we’ve seen reported in the news recently. So responsible AI is a major concern.

Predictive policing supporting proactive policing and risk prevention in public transport environments.

Q6. What’s more risky: giving AI too much autonomy, or never redesigning work around it at all?

Ben: If a business doesn’t adopt AI, employees will do it anyway through shadow AI – in other words, employees will use generative AI tools that have not been authorised by internal IT, like ChatGPT, for example. This brings its own risks. We’re at an inflection point, and competitors moving faster will create a gap that’s hard to close. You risk falling behind in efficiency and price competitiveness.

 

David: Pandora’s box is already open. To Ben’s point, McKinsey reports that 91% of employees are using generative AI for work, but only 13% of companies have successfully implemented multiple AI use cases. So not adopting AI is more dangerous than embracing it, as long as you embrace it strategically which is what Telefónica Tech helps organisations achieve. It’s about using guardrails, not avoidance – avoiding it just slows transformation and increases long‑term risk. (We’re covering this at an upcoming webinar: Governing AI Without Killing the Momentum.)

How to start using agentic AI

Q7. What mindset shift do leaders need to make to benefit from agentic AI?

David: In my opinion, leaders need to embrace continuous improvement. As we’ve been discussing here, AI does require a mindset change and is not a one-time thing. Start early, learn, iterate, and recognise that value comes from many small gains that compound over time.

 

Ben: I would say that foundations matter. If your data, permissions and estate aren’t in order, AI will expose those gaps. Companies often struggle not because AI is hard, but because their data culture and technology foundations aren’t ready. Cutting through hype and fixing basics is crucial, and this is, again, something we do often for Telefónica Tech customers.

Q8. For a team just beginning this journey, what’s the first meaningful step they should take?

Ben: Speak to Telefónica Tech, of course! Jokes aside, don’t try to ‘do AI everywhere’. Pick one process that costs significant time or money and fix that first. That thin slice becomes the reference case that proves value and builds momentum.

 

David: And make sure you’re involving all levels of people in that exercise – not just management. The people closest to the work know which repetitive tasks drain time. Bringing them in early surfaces real opportunities and aligns everyone on the value of agents. A great way to do this is to run agentic AI events such as Agentathons, Promptathons and Hackathons. We regularly host these at Telefónica Tech, and we help customers run their own.

 

Agentic AI is more attainable than you think

Q9. If there’s one thing you want readers to understand and act on after reading this, what is it?

Ben: To make the most of AI, whether using it in the agentic sense or another, my parting message is: start by getting the foundations right. Without good data and governance, AI can cause problems – like exposing sensitive information. But with the right groundwork, AI delivers transformative value, safely and at scale.

 

David: Additionally, start early and don’t assume AI is out of reach. Especially when working with a partner like Telefónica Tech, most organisations are closer than they think to benefiting from agentic AI. The biggest barrier is hesitation.

 

TAKE THE FIRST STEP

Bringing Agentic AI into your business processes

Clearly, the question is no longer “should we use AI?” Going forward, organisations must take a holistic view to implement AI, including agentic AI, into strategy and into the everyday – starting with high impact use cases to demonstrate value. Business doing this will pull ahead towards the frontier firm vision, innovating to better serve customers and improve the colleague experience.

 

To this end, Telefónica Tech’s Prism Framework helps organisations take a step back and get a clear view of where AI adds the most value in your business processes. This methodical approach goes beyond identifying quick wins, to delivering a solid foundation for secure, governed and scalable agentic AI adoption.

 

👉 What’s AI enabling in your industry, today? Discover how we’re working with your peers get inspiration on what you can achieve with our AI Use Case Library.

Learn More about Prism: Transform business processes with agentic AI

The Prism Framework

About the author

Related posts

news-telefonica-tech-uk (3)
Article
Agentic AI and The “Frontier Firm” Vision

By 2026, humans will oversee networks of autonomous agents handling structured and analytical tasks, transforming data management

Find out more
Healthcare AI Agent Use Cases
Article
Microsoft Copilot in Healthcare: 8 AI Agent Use Cases

Discover use cases of Microsoft Copilot in Healthcare, based on AI agents NHS Trusts built at our recent Healthcare Agentathon.

Find out more
Agentic AI: Prism Framework

Embark on your journey to the Frontier with Telefónica Tech's methodical Prism Framework for adopting agentic AI into business processes.

Find out more