Telcos already watch their cloud opex spend closely and GenAI may be about to make matters worse
Cloud providers have recently been espousing FinOps, which combines financial management with cloud operations to help organisations better understand their cloud spending.
The message lands well with enterprises, but for telcos grappling with launching new services while onboarding technologies, fully integrating with cloud providers opens them up to new cost challenges which generative AI is set to exacerbate.
Amdocs group president, technology, and head of strategy Anthony Goonetilleke highlights the risk telcos face when moving to cloud, suggesting that without FinOps: “it’s like the gift that keeps on giving for the cloud providers”.
“Telcos have run data centres and depreciated them with their costs going down,” he said. “Suddenly move your workloads to the cloud and you develop an ongoing opex cost.”
“It’s so easy to spin up a workload in the cloud. You get these developers suddenly going, ‘oh, yeah, I’ll spin up there…your invoice comes in at the end of the month and you’re like ‘what happened?’ – this is why telcos are looking at hybrid cloud even while embracing public cloud,” he told Mobile Europe.
IT cloud is not telco cloud
While public IT cloud services are well developed, the network side of public cloud is less so. Goonetilleke pointed to moves like Microsoft’s acquisition of Affirmed, plus AT&T moving its 5G core network into Azure, as clear milestones for the cloud providers in understanding how to shift packet core workloads to the cloud. Amdocs is a key partner in Azure Operator Nexus, which he likens to a telco operating system in the cloud.
Goonetilleke emphasises it is still early days as everybody still needs to figure out issues like GDPR and where workloads should be stored. “Are your data centres in-country, workloads in-country and what do you do about redundancy?” he questioned.
“There’s huge benefits trying to normalise your workloads,” he said. “Where we are right now is trying to figure out how much of that network stuff, how much of that edge cloud stuff, needs to stay on-prem versus being on a public cloud.”
He added that more public cloud providers will end up offering on-prem data centre solutions for telcos as a result, citing Oracle as one already with an offer.
Between a rock and a hard place
Goonetilleke said telcos are sometimes torn between devising an overarching cloud strategy, which may take years, and solving customers’ most burning use cases using cloud. The first, he says, delays monetisation while the second creates a mess when it comes to architecture, duplication and costs.
“The best approach is a hybrid of the two. It’s about tying your business imperatives with the right architecture, but not allowing a utopian architecture to rule a pragmatic approach of business results,” he said.
Amdocs is seeing its customers taking diverse approaches to this. For example, T-Mobile US is laser-focused on fixed wireless for broadband, rolling it out rapidly. Others like Comcast are looking at enterprise services.
Goonetilleke contrasts consumer experiences buying telco services with current enterprise provisioning. “I can almost get same day delivery of a device I ordered if I order it early enough in the morning and have it delivered to my house activated with eSIM. Switch it on you’re good to go,” he said. “On the enterprise side, it may take [up to] 44 days.”
Generative AI enters the fray
Amdocs recently launched its GenAI-powered Cloud Management Platform which, already integrated with public cloud providers, features a set of genAI-powered solutions designed to accelerate automation.
Goonetilleke believes the platform is a game-changer, addressing many of the telco cloud issues service providers are grappling with. For starters it securely manages thousands of applications in hybrid and multi-cloud environments under central governance and compliance policies.
At its heart is Amdocs’ GenAI framework, amAIz, which lets the vendor’s teams working on telco projects access and reuse hundreds of test automations, ready-made landing zones and continuous development pipelines. This means rather than relying on consumer-grade GenAI – with all the pitfalls hallucinations may bring – the telcos get enterprise-grade practices and security policies telcos can adopt quickly and feel safe doing so.
He acknowledges the issues around hallucinations, emergent behaviour, bias and observability – adding that telcos must be even more careful because, as regulated entities, they have to know how each decision was arrived at. Telcos are also going to need to consider AI’s energy consumption.
“One of the one of the issues we found around genAI is you can give it the same inputs 10 times it could generate 10 different results, which is problematic,” he said. “If I’m giving an offer to a customer based on certain criteria, what happens if you know two different customers asked for the same thing and you’re given two different pricing points?”
“The amAIz platform that we created starts to take these real-life things into consideration,” he said.
If managed properly, he said, GenAI can transform business processes. Goonetilleke highlighted a pilot where an enterprise salesperson was speaking to a customer about a new network build and, while they were speaking, GenAI produced 75% of the framework for the proposal in real-time – using access to inventory, previous proposals and existing topology.
“Even if it could do 100%, I wouldn’t do it because it needs to be verified and validated to make sure there’s no hallucinations are emergent behaviour,” he said. “But that 44 days can be cut down to a week.”
“We’re also seeing efficiencies for software development lifecycles. We believe that there’s double digit gains when it comes to doing stuff,” he added.
In addition, marketers can use genAI to produce a range of fully-templated, segmented offers quickly where they can choose one and customise – all in natural language.
The takeaway is that GenAI alone doesn’t fix everything. There are plenty of solid machine language and AI models that telcos are already using to varying degrees of success and he said the vendor’s platform can manage these holistically, automating whole business processes instead of single tasks – critical in hybrid cloud environments.
The cost of GenAI for telcos
For telcos, the cost of cloud services is going to be compounded by the inevitable cost of utilising AI-based apps, given how expensive and power-intensive LLMs are. If agents across a telco are implementing GenAI applications and then using them hundreds of times, what happens when these apps are all paid for?
Goonetilleke suggests that is why a cloud management platform is so important and it can be used to manage GenAI use while the industry settles down and its true cost is better understood.
“There’s going to be some type of normalisation. AI, ML and automation are all still good,” he said. “Telcos haven’t even exhausted RPA yet. With automation, you can test it, it’s replicable and there’s still great value. GenAI is not the panacea to solve everything else.”
Monetising 5G
Beyond managing cloud and AI costs, telcos need to up their billing game. Goonetilleke said the vendor was seeing good momentum with its new 5G Charging efforts which does away with siloed BSS platforms for each service launched in favour of a scalable cloud-native charging platform.
Amdocs currently has more than 20 cloud-based charging projects in production across a range of public, private, and hybrid environments, including at leading CSPs in North America and a tier-1 service provider in Europe.
In addition to cutting down the charging engine’s footprint with its latest version, Amdocs focused on improving its edge architecture capabilities, meaning telcos can rate and charge at the edge versus bringing all that data and information back to the central cloud.
Goonetilleke emphasised that unlike telco BSS providers, cloud providers are not used to the transactions per minute levels of real-time processing needed by operators – one global customer is expected to total more than 80 billion transactions a day – so the vendor has worked closely with AWS and Microsoft to generate that scale and run billing and rating in the cloud.
The end goal is automation
Amdocs’ acquisition of the service assurance business of TEOCO in May provided an insight into where the company thinks telcos will be heading with closed loop automation.
A decade ago, the average telco customer may have had a single SIM and service, but today that could be up to anywhere between 4-6 depending on devices, including intelligent homes, cars and smart meters. Telcos need to be able to assure each of these proactively through service quality management, fault management and performance management.
The vendor calls this new service assurance capability Helix and Goonetilleke said automation is the only way telcos will be able to manage the increasingly complex demands from customers.
“There’s a renaissance around the next generation OSS; how telcos automate their networks, end-to-end service orchestration and provisioning,” he said. “You can’t do it manually anymore; you cannot get to that level of [service] plurality without automating without providing assurance, without self-healing.”
As a result, he expects a lot more experimentation from telcos, citing AT&T RedCap and VMO2’s recent 5G in a box as examples. He also suggests the industry needs to collaborate more to develop new revenue streams.
“Roaming was probably the last example of successful cross industry implementation,” he said. “We need to look at how we monetise things better and I feel like it’s not a problem that one service provider can solve by themselves.”