Imagine a customer calling his operator and, before he says anything, the system already knows who he is, why he’s calling and what happened during his latest interactions. He doesn’t have to identify himself. He doesn’t have to repeat his story. He doesn’t have to browse a menu. He simply states what he needs and receives a helpful reply, in the right tone and with the full context.
If he needs to talk to a person, that person already has all the context. If he prefers to resolve it via a chat or the app, the conversation continues from where it left off. The channel is a choice, not an obstacle.
Now imagine the following evolution. The customer doesn’t even need to call. The AI agent has proactively detected an anomaly in his service and diagnosed the cause and it’s contacting him via a certified channel to inform him that the problem is already being resolved. It’s what we call invisible care: the best care is the kind that the customer doesn’t need to ask for.
That experience isn’t science fiction. This is what artificial intelligence allows us to do today and it’s what we’re building. Not as a layer of automation on top of existing processes, but as a different way of articulating our relationship with our customers.
However, for this to happen, we have to start in the right place.
The experience the customer deserves
Customers interact with their operator on dozens of occasions throughout the year. A query about a bill, a breakdown, a renewal or a change of plan. All too often, the friction is repeated upon every contact. The customer feels she has to retell her story, retrace the same steps and browse the same menus.
The expectation has changed. The customer expects her operator to know her, remember her context and be able to anticipate her needs. It isn’t an unreasonable expectation. It’s the experience she receives in other digital contexts every day.
The problem is that most customer care systems aren’t organised to deliver that experience yet. It isn’t a problem of willingness or talent. It’s an architectural problem.
Vision before technology
A transformation project often begins by choosing the technology and then looking for the problem to be resolved with it. It’s the fastest path to irrelevance.
The starting point isn’t the AI language model to be implemented. It’s the kind of experience to be delivered. The customer’s vision defines the architecture, not the other way around.
When it’s approached this way, some things become very clear. Network data and experience data have to converge into a common layer. This convergence is what enables us to offer the best possible digital experience, not only in the services that the customer subscribes to, but also in the support and care processes accompanying them. The goal is to ensure that the quality of the experience is consistent across all the contact points.
And there’s something else that matters: governing the experience as a product. With clear control, impact metrics and continuous iteration. This is what distinguishes a real transformation from a collection of disconnected initiatives; someone is responsible for the outcome, not just the delivery.
From chatbots to agents
For many years, customer service has functioned like a chest of drawers. The company keeps the solution to your problem in a drawer and puts a label on it, and you have to make the effort to find out which drawer it keeps in. IVR was just that. Chatbots, too; a slightly more sophisticated decision tree, but the same principle. If your problem matched those in one of the expected branches, you received a reply. If it didn’t, they put you through to a human. That’s no longer the case.
There are no more generic solutions hidden away in labelled drawers. The paradigm is different now; agents construct a personalised reply for your specific story. A telecommunications services portfolio is a living and fragmented organism (rates, devices, home services, IoT, business solutions, etc.) and agentic systems are the first ones that are capable of navigating this complexity to simplify it for you, personalising whatever’s general to your context and history.
But this capability requires alignment. An AI agent must be aligned with the company’s tone, values and principles. And this alignment has to be at the service of the experience, not the metrics. There’s an important difference between an agent who replies quickly and one who actually resolves the query. If we optimise things to complete interactions and the customer receives the same service as before, we haven’t made any progress. The goal is for the agent to resolve the problem, not just deal with the query.
Cooperation, not replacement
This is the most important point we need to clarify, because it lends itself to misunderstandings. AI isn’t meant to replace the customer care team. It’s there to enhance its capabilities.
The human agent arrives at each conversation having already summarised the customer’s context. He doesn’t have to search for information in five different systems before he can help. He already knows why the customer is calling, what’s happened before, what’s been tried, and what remains to be done. He can devote all his energy to what really matters: listening, interpreting and resolving.
When a query requires empathy, judgement or a complex decision, the AI agent intelligently scales and transfers all the accumulated context. The professional takes over without the customer having to repeat her problem. This detail, which may seem minor, is the difference between a frustrating experience and one that generates confidence.
And there’s an additional dimension; AI facilitates a care experience that’s truly omnichannel. The customer can initiate a process via a chat, continue on the phone and complete it on the app, without losing track of things or forgetting the context. The channel ceases to be a barrier and becomes a choice for the customer.
It doesn’t stay in the call centre. The field technician is able to ascertain the customer’s history before arriving for the visit. He knows what’s been tested, what’s failed and what the configuration is. He improves the resolution during the first contact, which is ultimately what the customer wants: to have the problem resolved immediately.
Measuring what matters
There’s a very common temptation in AI projects: to measure only the technical aspects. The model’s latency, the tokens consumed and the auto-resolve rate. These metrics are necessary but insufficient.
What really matters is measuring the impact on the customer experience and the business. The avoidable contact rate (the number of calls that shouldn’t have occurred). The actual resolution during the first contact (not the one the system indicates, but the one the customer perceives). The transactional NPS per moment of the journey (not an aggregate NPS that says nothing useful about the specific interaction).
These experience metrics have to co-exist with the technical metrics. But the relationship between the two isn’t neutral. Optimising the token consumption or response speed is legitimate and necessary. What cannot happen is any deviation of this optimisation from the real objective: to create a better experience. If we reduce the cost per interaction but the quality perceived by the customer decreases, we’ve won the wrong battle.
Model drift and hallucination detection are also essential. The technological layer that underpins the customer experience has to be governed with the same rigour as any other critical business asset. But with experience as the guiding principle.
Closer in every interaction
AI doesn’t transform the customer experience on its own. It does so when there’s a clear vision of the experience to be delivered and when there’s the determination to govern it as a product.
In telecommunications, a company’s distinctive feature isn’t just connectivity. It’s the quality of the relationship it builds around it. Having the best network is a must, but it’s also vital to provide a digital experience that lives up to expectations.
The goal is simple to enunciate but difficult to execute: there should be no more generic boxes separating the customer and the solution. The technology should adapt to each person’s story, not the other way around. Each interaction should be better than the previous one, because the system knows the customer, remembers the context and increases the human team’s ability to support her.
We’re back at the beginning. Imagine a customer calling her operator and not having to identify herself, repeat her story or find out which drawer her problem was stored in. She simply gets help. This is what we’re building. It isn’t a distant horizon. It’s our daily work.
