Telcos must take steps internally to become companies underpinned by AI. We look at implications for the organisation, its people and AI governance processes and tools to responsibly root this transition.

The implications for telcos on the road to an AI-native model

When considering what telcos need to become an AI-native organisation and what they need to do to support the rollout of responsible AI across the company, it is good to start with a view of what changes AI and machine learning (ML) will bring. Beyond the hype of potential new capabilities and fears of AI taking over the world is a general expectation that AI models will be deployed across the telco where it is the best technology to solve a problem and either provides bottom-line value or becomes part of broader telco goals such as the autonomous network and delivering new revenue streams.

We looked at a variety of expected changes in our 2023 research, which concluded that intelligence and automation would bring:

• New support for employees, including digital assistants/copilots deployed across the organisation and specifically for customer-facing teams or those working in complex areas such as the network. And new capabilities for customers, partners, and those working with telcos in the ecosystem.
• Intelligent automations.
• Gen AI-related functionality that provides either a creative (content generation) or dynamic (taking action such as generating an API) outcome.
• New paradigms including “machine as customer,” where machines purchase on behalf of humans.
• An increasing customer expectation that the experience will be personalised and straightforward.
• Long-term change in the business environment where new intelligence speeds up the business cycle and opens up markets that would have previously been difficult to enter.

Enter your details below to download an extract of the report

At the outset, many telcos had inflated expectations for the technology as part of their AI-native ambitions. In Figure 1, the blue line describes the corporate expectations and targets for AI deployment, starting with a goal of 100% of the organisation being automated with intelligence deployed where needed. The red line demonstrates the expected delivery against these expectations.

Successful deployment of telco AI

Source: Charlotte Patrick Research

Looking at the movement of the blue line over time:

• Initial AI hype drives a utopian vision for an AI-native telco.
• Enthusiasm rapidly defuses over the first year as reality sets in and the scale of the task becomes apparent.
• The next 5 to 10 years will be a push to cross the chasm – aligning expectations and reality.
• The final stage which moves towards maturity may well see expectations running slightly under what is possible, due to the appearance of new models that can support the development and deployment of new responsible AI.

The red line shows the actual deployment of AI being achieved as going through four main stages:

• Trials and early successes: The successful deployment of first models (where data is available and the problem relatively simple) with plenty of models not making it to production and only basic data and AI governance is in place.
• Steady speed of deployment: Progress depends on the availability of certain orchestrations (especially on the network), data availability and reworked processes clear of any issues; and a degree of sophistication within the organisation around creating/deploying models (such as governance, skill sets). Falling short in these areas will hold back the deployment of AI, as will questions about business models and the cost of deploying larger models.
• Deployment ramps up: A reduced cost of compute, the availability of more sophisticated models that overcome early model limitations (e.g. hallucinations), and an increasing comfort across the organisation with AI will see a ramp-up of deployment.
• Availability of new technologies: AI will increasingly be used in the creation and testing of AI models and help to provide sophisticated management of all models across the organisation. This will speed up the deployment of more AI models, potentially reducing the cost of creation and reducing the risk of deploying a model that develops an issue. At the point of full maturity of AI deployment, 10% of models which could be deployed will likely be left undeployed due to practical or cost issues.

Progress towards delivering on the high early expectations for AI is likely to be moderate into the medium term. However, the need for securing a solid organisational foundation for responsible AI deployment should be a short-term priority as deployment gains momentum.

Table of contents

• Executive Summary
  • The need for change
  • Recommendations
  • Next steps
• Introduction
• Creating a top-level vision for responsible AI and risk
  • Creating guiding principals
  • Risk management and mitigation
  • Measurement and monitoring
  • Example of a top-level vision for AI from Telefónica
• Developing an AI governance program
  • Running a successful AI governance program
  • Meeting expected AI regulations
  • Ethics committee and board-level oversight
  • AI governance framework
  • Example of an AI governance process from Telefónica
• Empowering organisation-wide people change
  • Building an in-house AI skill set
  • Cultural change
  • Leadership skills needed to build/maintain an AI culture
  • Training
• Deployment of supporting tools
• Conclusion
• Index

Related research

• Telco generative AI adoption tracker
• Finding value from AI, analytics and automation (A3) in the telco network
• Gen AI: Where should telcos start?
• Generative AI: Why and how telcos should skill up for success?

Charlotte Patrick

Charlotte Patrick

Associate Senior Analyst

Charlotte has 27 years of professional experience in strategy, marketing and finance. Most recently in the largest global technology analyst firm and previously two of the worlds largest global telecommunications companies. She is an electronics graduate and MBA with excellent business analysis, commercial and strategic planning skills.