‘Frontier’ telcos like Vodafone, AT&T, and Telefónica are deploying a whole lot of AI brokers throughout their operations, says Microsoft – to automate processes, speed up gross sales, and drive a brand new form of operational agility. It has a weeks–to-minutes use-case with Vodafone. However telcos are additionally slowing within the AI race due to three huge issues.
In sum – what to know:
Days to minutes – Vodafone is utilizing an AI agent to automate elements of its RFI/RFP course of, producing drafts in minutes as an alternative of the ten–30 days typical in lots of telcos.
Brokers in all places – Microsoft says ‘frontier’ corporations like AT&T and Telefónica have deployed a whole lot of brokers in all places from gross sales and finance to community ops.
Limitations to scale – Knowledge governance, enterprise expertise, and agent administration are rising as the important thing obstacles telcos should clear up to unlock the complete ROI of agentic AI.
Right here’s a very good enterprise case for agentic AI in telecoms – as deployed by Vodafone within the UK, as advised to RCR by Microsoft at MWC in Barcelona. It’s drawn from an increasing roster of Microsoft tasks with ‘frontier’ telcos, all-in on AI, which additionally (anecdotally) consists of AT&T and T-Cellular within the US and Telefónica in Spain. On paper, this stream of labor with Vodafone would possibly simply be probably the most highly effective and chic instance of how AI goes to work for telecoms. “In most telcos, you is likely to be speaking 10 working days, possibly upwards of 30. Whereas this takes minutes.”
So says Rick Lievano, chief know-how officer for telco {industry} purchasers at Microsoft. “Ten days is unacceptable in immediately’s world,” he says, within the again room of a busy stand in Barcelona. “These frontier corporations are the templates, for certain – having expanded outdoors of their contact facilities. As a result of AI isn’t just the contact heart anymore. AT&T has deployed over 400 brokers throughout its enterprise. Identical with Vodafone.” And so we hear about Microsoft’s work with Vodafone Enterprise to automate the UK agency’s RFI/RFP course of – a minimum of to the purpose a draft is on the desk.
Lievano says: “These product catalogs are exceedingly complicated, proper? Particularly B2B catalogs. This isn’t an e-commerce catalog with a couple of hundred gadgets; that is an enterprise telco catalog with tens of hundreds of them. As a result of any providing, with any tweak, turns into a brand new product. So they’re complicated, and these incoming RFIs/RFPs request this QoS and this SLA, and mandate a brand new fiber circuit or a wavelength, say, can’t go through this geography; and all of that must be chosen and documented by means of the response course of.
“What occurs immediately is that RFI/RFP sits in a queue till somebody picks it up, has a espresso, defines a response, goes to lunch, and picks it up once more three days later from one other queue of duties; after which they begin to reply. They evaluate a library of comparable requests, a set of knowledge-based articles, all of the procedures for methods to reply, all the correct checkpoints – just like the supervisor related to that account. So it’s a actually lengthy course of – from first receipt to first draft. Whereas Vodafone has an AI agent to choose all of that up, and begin the method autonomously.”
He explains: “It receives the request, understands the request, checks the system, notifies the stakeholders. And it doesn’t cease there. It has entry to the identical info: firm insurance policies, boilerplate textual content, previous responses, pricing info, product catalog. You title it, it has entry; and it produces a draft, 90-percent completed, in simply minutes – which exhibits what’s customary and what’s modified, and arms over to the account supervisor to see it out, to make sure compliance and pricing. The purpose is it isn’t beginning at zero anymore.”
And he provides: “It means you might be rather more agile – and agility is cash. You might be sooner than your opponents – and so, guess what, you might be higher positioned.”
AI FOR EVERYONE
Greater than that, you don’t even want a lot coaching, it appears. Low-code / no-code software program instruments (like Copilot Studio) supply a graphical “drag-and-drop” coding platform with a generative AI language interface so gross sales administrators and account managers, say, can dictate phrases on their phrases – to create such an RFI/RFP agent from scratch. It prioritizes area data each methods, in a parallel of the previous IT/OT dialogue – so the deep tech is abstracted and democratized (and popularized / scaled), and put to work in service of the specialist enterprise data.
“Brokers gained’t be created by engineers. Engineers know their very own area – methods to run a community, methods to write code. They don’t know the enterprise. The very best folks to create brokers are the folks doing the work, who want the assistance. Which is the thought with copilots – brokers, in multiples – in assist of pilots. The pilot is the gross sales rep – who is aware of the processes, paperwork, insurance policies. A software program engineer in a lab doesn’t know any of that. And now they’ve the instruments to create the agent, and select the way it’s triggered – whether or not it’s conversational, or hooked up to an inbox.”
There are a “million methods” to set off an agent, says Lievano. Once more: complete flex for complete alternative, throughout the limits the tech affords. The Vodafone case is a neat instance of an honest telco repair – “to drive an RFP to completion, or a minimum of to a primary draft”, he says, earlier than it enters a extra acquainted workflow – the place a human “reads it, confirms it, approves it, and strikes it to the following part, to yield a buyer supply”. As an anecdote, it additionally articulates a much bigger story in regards to the rise of AI brokers amongst these so-called ‘frontier’ corporations (over 400 at AT&T, plus no matter the remaining are doing).
The demos on Microsoft’s sales space (as demo’d to RCR) are extra expansive; there are many AI brokers, speaking collectively variously, to assist with hassle capturing, ticket administration, discipline operations – to drive autonomy within the community itself. Leivano displays: “The contact heart was the entrance door, and it was solely pure you’d see generative AI bloom there as a result of it’s a treasure trove of language. However that was the lab, the place the experiments started. The frontier telcos are increasing in all places – into gross sales, finance, HR, authorized, and into extra particular telco domains just like the community.”
The “messaging”, he says, is all in regards to the “ROI on intelligence and belief”. Which, very clearly, is a downstream dialog relative to the {industry}’s early call-centre experiments, and an indication, conversational or industrial, that the {industry} is maturing, and trying to stack AI use instances inside their operations. “Each telco is saying, ‘Hey, I’m investing, I’m driving, I’m creating pilots; so present me the ROI’. On the finish of the day, that’s what it’s about. We completely hear you; we perceive. And we’re offering a set of pointers based mostly on finest practices at these frontier telcos.”
However MWC is a telco present, in fact, and Microsoft has designated ‘frontier’ corporations in all places, in each {industry}. Telecoms has been good at AI (or good at pursuing AI efficiencies), historically; however it isn’t such a frontrunner, anymore. The agency has a analysis paper with analyst home IDC that polls 4,000 enterprise leaders from all sectors, and says ‘frontier’ corporations have put AI to work in seven completely different enterprise capabilities, on common, and two third (67 p.c) are monetizing “industry-specific” AI in a method or one other (58 p.c use “customized AI”, rising to 77 p.c by 2028).
AI AGENTS TRIPLE
In two years, the variety of frontier corporations utilizing agentic AI will “triple”, the report suggests (with out clearly giving a base-line begin). The pair has a best-practices presentation, as effectively – additionally value studying. Extra useful, Lievano says telcos have seen a 2.8-times return on their agentic AI investments thus far. However telcos are slowing, typically or comparatively, and each different {industry} is accelerating. Lievano feedback: “Initially, telcos had been a few of the first to undertake generative AI particularly, however… we don’t see that very same acceleration that we noticed two years in the past.”
On the stand, Microsoft says the par rating for community autonomy, early 2026, is someplace between two and three – out of 5 on the TM Discussion board scale (Ranges 1-5) for core-operational telco AI. It’s value noting that even claims from frontier corporations about Degree 4 capabilities, with a high-degree or ‘self-organizing’ and ‘self-repairing’ are largely siloed in sure operational capabilities, siloed in sure working corporations. Therefore, the significance in fact to speak about finest practices, additionally entering into top-line income era, on the likes of AT&T, T-Cellular, Telefónica, Vodafone.
Therefore why the message in regards to the related simplicity, in use and design, is essential, too. Leivano feedback: “They’re embedding AI throughout their companies. We would discuss to the networks staff, and we would discuss to the finance staff. Totally different conversations, similar basis – you design in the same method, regardless that the brokers are grounded and instructed in a different way, and related to completely different information and completely different instruments to do various things. However the structure could be very a lot the identical.” However there are obstacles, as effectively – three “compounding” ones, says Lievano.
The primary (and final; isn’t it at all times?) is information – or orchestration of it. “Folks say the issue is information silos, however that’s not likely the difficulty; it’s governance.” Telcos are wealthy in information, however such abundance masks a structural downside: the data will not be persistently outlined, trusted, or related throughout methods. Buyer information would possibly dwell in CRM platforms, billing platforms, product catalogs, and community databases – every with completely different identifiers and naming conventions. What one system calls a ‘buyer’, one other would possibly name a ‘subscriber’, says Lievano.
Functionally they discuss with the identical entity; structurally they’re handled as various things. The problem is much less about the place the information sits – whether or not within the cloud, whichever cloud, or on-premise; or more and more, distributed throughout the continuum with a number of distributors – and extra about how it’s found, understood, and managed. AI brokers, like each horizontal digital-change tech, solely work in addition to the data they’ll entry and belief. If the methods can not reconcile identities, schemas, and permissions, the agent can’t reliably sew collectively the proper image.
After which they lose accuracy and belief, and the home comes down. The reply, says Lievano, is to not centralize all the things in a single repository. “We’ve no expectation that each one the information strikes onto Azure,” he explains. As an alternative, the goal is to hook up with it wherever it lives, derive its construction and metadata, and construct a semantic cloud-SaaS layer (Microsoft Material) on high that reconciles relationships between datasets. In that sense, the governance problem is about creating an operational ontology – a shared understanding of what the information means and the way it relates.
AI GOVERNANCE
The second impediment is the abilities hole, mentioned already. Lievano provides: “Like we mentioned: anyone can create an agent. You don’t must be a developer. You don’t must have an Azure subscription. When you’ve got a Microsoft 365 subscription, which is 90 p.c of telcos, or extra, then you have already got entry to all the things you must create and host the agent – without having anything. So anyone can self-serve the brokers they want.” Once more, it isn’t a knee-jerk overhaul, about hiring AI techs, however about prioritizing in-house expertise to make the tech answer scale.
Though, there may be loads of irony in such a place, given the headcount discount in telecoms proper now, to a better or lesser diploma enabled by AI-geared effectivity drives. However that may be a separate dialogue. For Microsoft, the pitch is that low-code / no-code instruments are simply accessible and simply usable for operational groups to embed data and construct brokers. Which leads on to the third problem: governance once more – however this time for the brokers themselves. That is what Cisco talked about at MWC, additionally, and the place it’s in search of to hold its hat within the new AI gross sales discourse.
Lievano says: “Proper now, IT teams are horrified that each one these brokers are being unleashed throughout their enterprise methods and networks. They don’t have any visibility into what they’re doing.” Questions shortly multiply. How are they recognized? What can they entry? What are they utilizing? What are they costing? With out management, brokers will swarm and riot, successfully; with out correct governance, their administration might be inconceivable. Costly AI will do extra hurt than good, and belief might be eroded. That form of message.
Microsoft’s reply is an rising administration framework (Microsoft Agent 365; in preview now, for launch “quickly”), a centralized catalog the place enterprise brokers might be registered, monitored, and managed no matter the place they originate. IT directors can assign identities and permissions, observe exercise, and handle prices – whereas nonetheless giving reign for the broader group to create and deploy brokers as required. “That was the lacking hyperlink to allow us to go to the following stage of the agentic revolution the place we now have hundreds of brokers being deployed.”
These frontier telcos, which Microsoft has anointed, have been grappling with this downside, Lievano says. Many are constructing their very own governance frameworks because the variety of brokers grows into the a whole lot. The goal now’s to standardize that management layer so corporations can scale the mannequin safely. ““We’re saying, ‘No, you don’t want to try this; we’re going to offer the potential.’ We’re working with many telcos which have tried to handle the issue, and brought their suggestions into the event of Agent365; they’re amongst our first clients.”
Broadly, the enterprise dialog about AI has matured, says Lievano. “It’s not likely about whether or not try to be creating brokers anymore. That query has been answered. The dialogue now’s: how do I handle them, how do I management them, and the way do I be sure they’re delivering worth?” Which, ultimately, takes us again to Vodafone’s RFP instance. The tech is spectacular (minutes as an alternative of weeks), however the lesson is structural. If operators can clear up the governance questions, the enterprise answer is definitely fairly easy: agility. Which is cash – like the person says.
