How to turn AI hype into a practical reality

Author: Dean Smith, COO

Across the telecommunications industry, operators are under immense pressure to move quickly on AI. However, the primary risk is not being left behind, it is deploying the technology without clear use cases or reliable data.

Successful AI implementation depends on getting the fundamentals right first, including clean data, clear ownership, and measurable business objectives. Rather than attempting to become an AI-native organisation overnight, operators should focus on where AI can deliver measurable value today. A pragmatic approach prioritises meaningful operational improvements over broad transformation programmes with unclear outcomes.

A less discussed challenge to this pragmatic approach is cost control. While AI is often seen as a way to save money, many applications run on token-based consumption models. Without strict governance, these costs can quickly exceed the organisation’s budget. This financial risk highlights a much wider issue across the industry: the gap between ambition and implementation.

 

The gap between ambition and implementation

The industry is currently navigating a period where AI is often overpromised and framed as a sudden, transformational leap. While the potential for AI to address challenges such as tight margins and operational complexity is genuine, there is a risk of pushing the technology too far too soon. Many operators are still managing legacy systems, regulatory constraints, and siloed data, all of which can impede the progress of AI projects if not addressed upfront. Not to mention the inevitable workforce and cultural implications.

Research from McKinsey highlights the common pitfalls of rushed deployments:

“It turned out that the early adopters in the pilot were more enthusiastic than the rank-and-file users when the program was rolled out throughout the organization. Managers also cited poor communication between tech developers and business teams, incomplete or inconsistent data foundations, and limited change management to drive adoption.”

If the industry does not ground the promise of AI in operational reality, it risks repeating the cycles of previous technologies that failed to fully materialise, such as O-RAN or early iterations of robotic process automation (RPA). Rushed deployment without the necessary groundwork leads to wasted investment, low adoption rates, and long-term technical debt.

 

A framework for practical AI in telecom

To integrate AI in a way that evolves the business effectively, operators should follow a disciplined implementation process. This involves understanding where different components of AI are best suited and prioritising programme management.

Strategic planning and design upfront

The conception phase and initial programme planning are the most critical stages of any AI project. A technical course is difficult and expensive to alter once implementation begins, so it’s essential to go through thorough concept, design, and project planning first.

This stage should focus on:

• Creating clear definitions of what success looks like for each specific use case, including what ‘flavour’ of AI you are hoping to implement.
• Finding leaner ways to execute processes, ensuring that the chosen solution serves the interests of the business rather than the vendor.
• Planning integration into the existing workflow, rather than deploying AI as an isolated tool that adds to the complexity of the operational environment.

Independent expertise is particularly valuable here, as it provides an objective view of legacy systems and data structures that the vendor may not fully understand. By building common sense into the plan from the beginning, operators can avoid the trap of pursuing technology for its own sake.

To dive into this a little deeper, take a look at our consultancy services and how we help you plan for project success.

Synchronisation and multi-vendor integration

Most telcos operate in a multi-vendor environment, which creates significant challenges for data consistency.

A practical approach to AI requires simplifying and standardising reference data so that it can be pulled into different platforms, such as rating environments, billing systems, and customer care portals. By keeping data clean and ensuring it’s synchronised across these disparate systems, operators create a unified automation layer.

Operational controls and defined boundaries

Effective AI deployment requires an environment with clear boundaries and human oversight. While it’s possible to automate a vast number of tasks, there will always be exceptions that require human processing.

Establishing a system of freedom within fences allows for automation to scale while ensuring that AI operates within its intended scope, and humans remain in the loop for strategic concepting and decision making.

YOU MAY ALSO LIKE: Reference data integrity and AI: why automation needs clean definitions

The human element and change management

It’s essential to bring the workforce along on the journey of implementing AI. By using the on-the-ground expertise of your team, you can both understand where AI can make the greatest impact and build confidence in the tools.

When employees understand how AI can help them rather than replace them, you’re more likely to see higher adoption rates and commercial impact. This bottom-up perspective ensures that the AI addresses meaningful, everyday challenges faced by the staff.

Prioritising high-impact use cases

Instead of attempting a blanket transformation, operators should focus on defined operational challenges where margins are under pressure or where large-scale processes offer the biggest cost-saving opportunities.

The goal is to identify the 80/20 gains – where 80% of the benefit can be achieved with 20% of the effort – by selecting specific use cases with a measurable return on investment.

Practical examples include:

• Process efficiency: Automating high-volume manual tasks alongside cross-platform data and engagement for more timely and effective results.
• Test data management: Using AI to improve the accuracy and availability of data for system testing.
• Fraud detection: Supporting fraud teams by using AI to prioritise suspicious activity, collate and associate case information, and infer risk and actions allowing human experts to make the final decisions based on outcomes.
• Predictive maintenance: Assisting network engineers by identifying potential faults and root causes before they lead to service outages.

By scaling gradually based on proven impact, operators can learn from each implementation and refine their approach. This prevents the business from overextending its resources on projects that do not deliver clear value.

What’s promising about this latest wave of automation, is that the cognitive abilities of AI allow us to automate tasks that were previously too bespoke or complex for previous iterations of technologies like RPA. This evolution, that includes agentic AI, can improve the quality of processes and maximise cost saving potential.

 

Data as the critical foundation

The integrity of an operator’s data is the single most important factor in the success of an AI strategy. If reference data points are inaccurate or inconsistent, AI will simply amplify these errors, leading to flawed outputs and operational failures. Core data governance and integrity must be established as a prerequisite for any AI project.

Common failure points include incomplete data foundations, poor communication between technical and business teams, and a lack of clear ownership of data and related processes. Before scaling AI, telcos should prioritise data integrity to ensure quality and validate for use in AI-driven scaling. Establishing a clear data governance framework and reference data management practices will ensure that the AI has a reliable source of information to work from.

Are you ready to strengthen your data foundations?

Download our latest whitepaper, ‘Fixing your Foundations’, for a detailed look at reference data management best practices and how to prepare your organisation for successful AI implementation.

[Download now ->]

 

Becoming AI-ready

The objective for telecommunications operators should not be to become ‘AI-native’ overnight. Rather, the goal is to become AI-capable in a disciplined and commercially grounded way. While vendors may promise end-to-end transformation, success is determined by the groundwork and the discipline applied to the implementation.

The operators who benefit most from AI will be those who implement it with a strong foundation of data integrity. By focusing on practical use cases, maintaining human oversight, and ensuring that data is accurate and synchronised, telcos can achieve meaningful improvements in efficiency and cost management.

AI is a powerful tool for evolution, provided it is grounded in reality.

Contact our team to talk about how you can prepare your organisation for a practical and successful AI deployment.