From AI hype to autonomous impact
How CSPs are turning intelligence into operational results
Discover how generative AI and agentic AI are transforming communications service providers (CSPs). Based on Omdia global surveys and executive insights, this e-book explores AI's impact on efficiency, ROI, and the future of telecom operations.
From AI hype to autonomous impact: How CSPs are turning intelligence into operational results
Since the debut of ChatGPT in November 2022, interest in generative AI (GenAI) skyrocketed, with enterprises of all types investigating how to leverage its capabilities. Communications service providers (CSPs) are on perennial digital transformation journeys, so they were keen to see how GenAI could transform their network and business operations. A few years later, agentic AI burst on the scene, offering the promise of autonomous operations – something many CSPs have been working towards as they aim to improve efficiency in the name of an improved customer experience.
As telecom industry watchers, Omdia analysts are always looking for evidence of the impact of AI on CSP operations. While it may be difficult to draw a straight line between deployment of more intelligent chatbots and increases in revenue, Omdia aims to study what benefits and outcomes CSPs are achieving today and how AI is changing ROI expectations.
Omdia developed an online survey directed at CSP executives with direct knowledge of their company’s AI efforts. Receiving more than 60 validated responses from CSPs around the world, Omdia obtained a high-level view of where CSPs are deploying various forms of AI today, and more importantly what outcomes they are achieving. To supplement these findings, with the valued support of the organizers of the FutureNet World conference, Omdia spoke directly to multiple C-level executives driving AI-related projects at leading CSPs. Their candid and extensive input provided more nuance to the simple question of “How does AI impact operating expenditure (opex)?”.
Don’t wait until everything is certain. Find a small use case that causes big headaches, and see how AI - in any of its forms - can help. With any new technology, the best practice is to start small and use that learning to take the next step. You don’t need to move quickly. Just move.
Remember the customer. It’s easy to get caught up in the breathless AI hype, but your responsibility is first and foremost to your customers. If you use AI to make their experience more reliable and enjoyable, any opex savings will become secondary to revenue gains.
Be steadfast on governance. CSPs historically have guarded the networks in their care with the utmost diligence. With AI technologies evolving so quickly, now is not the time to let down your guard. Agentic AI can, and should, be part of your roadmap - but you must make sure guardrails are in place to ensure they do only what you ask them to do.
Make it clear to your CSP customers how investments to improve opex in network and business operations will lead to an improved customer experience. Show them how AI can help monetize that improved experience so they can continue investing in a virtuous cycle.
Help agents shine by providing clear and solid guardrails. Autonomous network operations are at the top of nearly every CSP wish list. Identity and access management for agents needs to be robust and auditable. As agents prove themselves capable of operating within their limits, CSPs will become more comfortable turning over more functions and processes.
Build for interoperability and telco‑grade ops. Offer open APIs and TM Forum-aligned schemas, ship connectors for major OSS/BSS, ITSM and NMS tools, and support hybrid/on‑prem deployments including air‑gapped environments. Ensure high availability, observability, and zero‑trust security as first principles.
Somewhat surprisingly, our survey shows that predictive, generative and agentic AI are nearly equally deployed in production at CSPs to support network or business operations. The following definitions were provided.
Predictive AI: AI which analyzes new data using algorithms or pattern recognition
Generative AI: AI which generates new data points based on relationships in training data
Agentic AI: AI which reasons and performs complex tasks autonomously
What is not captured, of course, is the extent to which the various types of AI have been deployed. We would expect that predictive AI has a much larger footprint than agentic AI given the different maturity levels of the technologies. Still, as telecom industry watchers, Omdia is pleased to see the scales tipping towards more CSPs trying AI than fewer.
Asking about CSPs’ use of AI requires much more specificity than headlines would suggest. Is the question asking about the use of GenAI coding assistants to make software development teams more effective? Is it giving employees access to Co-Pilot so all departments can achieve efficiency gains? Or is the question how GenAI is transforming how CSPs operate their networks? Most large CSPs are doing some combination of all of these and achieving similar types of results.
The difference comes in creating the business case upfront. Traditionally, any spending on hardware or software needs to show a clear return on investment. With AI, however, its diffuse impact is much harder to anticipate, and therefore put an exact dollar amount on the potential benefit. That said, CSPs in this study, and others Omdia has conducted, revealed that they know AI tools can make their organizations more efficient, so even if one cannot pinpoint an exact figure, the investments are able to proceed. However, when it comes to use cases targeting a particular process, the business case is just as important as ever. Omdia has found that the usual key performance indicators (KPIs) like mean-time-to-repair (MTTR) , net promoter score (NPS), call resolution rates, etc. are still used when evaluating AI solutions.
These metrics are collected during the trial phase and included when making the business case for further investment.
In our survey, we asked which factors are most responsible for reducing opex in various KPIs/metrics. The chart shows that process efficiency was rated as the most responsible for all the KPIs except lower call handling times, where the need for fewer employees was highest. For all the metrics, the need for fewer employees was ranked higher than the need for fewer contractors.
However, when we spoke with CSPs, we learned that some had moved the discussion away from opex improvements - tying everything to improving the customer experience. One CIO explained that there is a risk that optimizing for something like network configuration time, could end up negatively impacting the customer experience. This insight led his organization to make customer experience the metric – understanding that improving operational efficiency by its nature should always improve the customer experience.
CSPs are vast and complex businesses, with numerous cost centers. It’s unsurprising, therefore, that AI’s impact on opex varies by function. We asked respondents both how much AI will reduce opex in different categories and when those effects will begin. The chart shows the results for three of the categories, network operations, general and admin, and utilities. Across all three, ‘moderate’ was the most common response.
In network operations, 37% of respondents expect a ‘high’ opex reduction from AI, and 36% say the impact is already happening, with another 36% expecting it within a year. AI can address network operation cost-drivers. For example:
Predictive assurance and anomaly detection can reduce incidents and truck rolls
Automated ticket triaging can accelerate MTTR
Agentic AI may be able to transform currently employee-led processes to increase automation and speed up operational processes
These capabilities translate into lower field labor costs, fewer escalations, and improved SLA adherence. The caution shown in some responses likely reflects integration complexity across multi-vendor estates and the need for robust, high‑quality data and MLOps to support models in production.
General and admin presents a different profile. AI can help eliminate repetitive activities in finance, HR, procurement and legal workflows. While this can reduce labor requirements, cost benefits also come from fewer errors, faster processes and redeploying people towards high-value activities which can improve employee experience and retention too. A challenge is that AI benefits are often distributed across many processes and teams, making it harder to isolate and measure.
Our respondents thought AI would have the least impact on utilities (which is primarily spending on energy) with only 20% answering 'high' and 31% answering 'minimal'. The RAN accounts for up to 75% of a mobile operator's electricity consumption and AI can improve RAN energy efficiency (sleep modes, traffic-aware shutdown) and optimize cooling. However, impact is constrained by hardware capabilities, vendor feature support, telemetry depth, and modernization cycles.
Building and operating networks in cloud environments is not a strict prerequisite for leveraging AI capabilities across different functions. However, our discussions with CSPs reveal that AI capabilities, particularly generative and predictive AI, are often more effective in cloud environments. This is because cloud-native systems are built on open, standardized interfaces (APIs), which allow AI tools to seamlessly access, query, and act upon data across various systems. In contrast, legacy on-premises architectures, with their proprietary and closed nature, often hinder such integrations.
Notably, 41% of our survey respondents believe AI can have a significant impact on reducing operating expenses for IT systems in cloud environments. A CIO we interviewed emphasized the role of practices like Infrastructure as Code (IaC) in transforming cloud resource management. AI is increasingly being used for dynamic scaling, creating new environments on demand, and enabling autonomous actions. These are capabilities that are difficult to achieve in legacy systems.
In an increasingly complex world, clarity and decisiveness are just enough to compete. But to get ahead, you need a partner who sees your market, knows your customers, and understands the intricacies of your business.
Take the next steps with Omdia.
With expertise around the world and comprehensive coverage of the tech domains critical to your business, we give you the breadth to see the big picture and the depth to access decisively.
Our clients also enjoy direct access to our analysts, helping you navigate your competitive landscape, shape your go-to-market strategies, and open new paths forward.
Breadth of research: 42+ technology areas, 200+ markets, 3,000+ topics covered. 12,000 tracked companies, 24,000 channel partner community.
Depth of analysis: 200 million+ data points, 100,000 primary research respondents p/a, 1mil+ daily intent signals, 50m+ audience, 3,000 reports annually, 250+ forecasts annually, 2,500+ consulting projects, 5,500+ deliverables annually, 100+ economic and technical validations p/a.
Markets we cover:
Ready to leverage our world-class intelligence and advisory services?
Fill out the form on the final page of this e-book, and a member of our team will be in touch.
