An AI-first operating model: A new imperative for telcos

Insights

• Many telcos are embedding generative AI into their existing connectivity portfolio, achieving good returns but minimal value-add.
• To turn generative AI capabilities into new products and solutions that provide significant business growth opportunities, telcos need an AI-ready operating model.
• The model proposed here has human-centricity at the core, is guided by transformation management, and is composed of the six pillars of: AI product management, governance, intelligent experience, enterprise AI architecture, platform ecosystem and AI factory.

In our companion article to this paper, we described how telcos can become “tech-cos” by utilizing AI technologies to fundamentally change their product strategy. A quick way in is to integrate generative AI into the existing connectivity portfolio, but they can go one better by providing value-added services to other industries, including fleet management and customer incident management solutions.

We also explained that to do this, telcos must first get their data estate in order, must decide on build versus buy versus partner strategies, and ensure executive sponsorship to make sure AI initiatives can scale appropriately.

Another huge aspect of institutionalizing generative AI at scale is the need for an AI-first operating model.

Key insights from companion paper

“The telco of the future: Generative AI as a force multiplier in product strategy”

• Generative AI is driving significant disruption across the telco industry.
• But to get the business benefit, telcos will have to look beyond low-hanging fruit use cases, and deal with more complex, capability-driven initiatives that deliver long-term value.
• Firms that want to go a step further will embed this technology into product strategy; how much value-add they can achieve depends on their data and AI readiness.
• For firms that don’t have data science talent, partnering is a good strategy, with partners bringing the AI capabilities and telcos the domain-specific know-how.
• Key imperatives to transform from telco to tech-co include identifying high-impact use cases and breaking complex projects into short, sharp sprints that deliver exponential business benefit.

AT&T, for example, puts engineering excellence and transformation management at the heart of its operating model. Software engineers, data scientists, and thousands of citizen developers build enterprise applications that are changing the telco behemoth to an AI-first nimble tech-co. With this operating model in place, AT&T is using generative AI to optimize its networks, upgrade legacy software, enhance contact centers, and upskill employees around the three pillars of human-centricity, responsible AI, and secure/ethical by design.

However, to turn capabilities such as software development, customer service, and generative AI-powered insights into AI-ready products and solutions, firms should have human-centricity at the core. This AI-talent-ready model should sit at the center of a six-pronged operating model (AI product management; governance; intelligent experience; enterprise AI architecture; platform ecosystem; AI factory) that also includes transformation management as a foundational building block.

Figure 1. An AI-first operating model to transition from telco to tech-co

Human-centric focus

AI should augment rather than replaces employees. It does this by unlocking access to the intelligence within all organizational assets, including artifacts, systems and humans. A tech-co will also have a skilled workforce proficient in AI and data. To get ahead, tech-cos need both creator community and consumer community AI skills. Creators include data scientists, econometrists, machine learning engineers, and other AI experts. Vodafone CIO Ulrich Irnich says that 42% of the company workforce are software engineers, which will increase to 50% by 2025.

Additionally, prompt engineering is a key skill required. Anthropic, Google’s $300 million investment, is hiring prompt engineers at salaries up to $335,000 a year, underscoring the value placed on prompt engineering and its role as a crucial skillset. Even newer jobs will span data, engineering, and design, and must incorporate AI-driven learning paths to ensure readiness for the future.

AI product management

The future telco will be product-centric. This is a mindset and culture shift from AI projects to AI products and services. It emphasizes developing production-ready, client-facing AI solutions. Tech-cos will differentiate by organizing teams around “flow of value” by reimagining capabilities as products aligned to customer journeys and associated value streams (Figure 2). For example, one value stream could be software engineering. Teams (or pods) would need to be cross-functional, encompassing application development, maintenance, testing, modernization, architecture and design, and new services (including personalized AI assistants). By organizing in this way, tech-cos eliminate internal team siloes and can more readily take advantage of the platform ecosystem – an important element in the AI-first organization. Indeed, telco CEOs that take a bird’s eye view of their businesses and go both product- and platform-first will be able to sync investments across 5G, cloud, and IoT.

Further, this mindset and structure leads to faster product launches, more innovations, and system stability.

Figure 2. An example of a product-centric operating model for telco firms

Governance

Adequate governance is needed to ensure generative AI models work as expected and with appropriate guardrails. It also includes the concept of responsible by design, which ensures model and data safeguards are instituted at the beginning of an AI project. Responsible AI also ensures that products are ethical, secure, and lawful, with thought given to IP leakage and the regulatory ramifications of a new product. This is becoming ever more important in Europe with the introduction of the EU AI Act.

By getting this lever right, tech-cos can safely scale initiatives such as intelligent customer segmentation and network optimization, while building trust, ensuring sustainability, and mitigating risks throughout the full AI product lifecycle.

Intelligent experience

A big part of AI-first is embedding generative AI into all processes and experiences, including chatbots, network security testing, and interactive engineer support. This is a strategic shift to streamline operations, use AI to enhance decision making, and deliver superior customer experiences through insight generation and content intelligence tools.

“There’s no longer any distinction between business strategy and the design of the user experience,” says Bridget Van Kralingen, senior vice-president of a global business services provider.

To get a head start, telcos can identify processes and experiences that will benefit from AI-led microchanges, then use AI assets to deliver the reimagined process. The Infosys AI Store provides more than 12,000 AI use cases, and more than 150 pre-trained models, 100 datasets, and 50 AI templates to unlock process value at scale. The caveat is that no matter how much AI is used in the reimagined intelligent experience, a human in the loop is still necessary to comply with regulations, and ensure trust, transparency and explainability.

Enterprise AI architecture

A generative AI reference architecture creates an enterprise level blueprint for generative AI use at scale. This platform approach ensures changing AI frameworks, a fast-moving product and vendor landscape, and evolving telecommunications regulations. As we wrote about in our companion article, models change all the time, and it is necessary to have an abstraction layer that sits above the model so that tech-cos can swap in new models when they are needed, or when performance degradation is a problem.

This generative AI blueprint should be futuristic, enable democratized development, and be both self-governed and scalable. Futuristic means that each tech-co divides its platform into loosely defined layers with well-defined responsibilities and enables unified experiences for various roles and corresponding tools to increase adoption and productivity. As mentioned, product teams should also be able to get what they need in a self-service manner, giving each individual the ability to consume, and contribute to, AI knowledge without barriers. Scalability means going cloud-native, enabling the onboarding of new AI technology at any time, and optimizing AI components and infrastructure at enterprise scale.

Partner ecosystem

Partnering is important when there isn’t requisite AI IP in-house. After considering go-to-market solutions, telcos can select strategic partnerships where the partner brings the AI capability and the telco brings expertise in IoT and domain-specific know-how. This ecosystem encompasses a wide range of stakeholders, including AI technology providers, research institutions, startups, and industry consortia.

Also, partnering is important to ensure data and AI readiness. As data transcends boundaries beyond the purview of single telcos, companies are evolving into ecosystems to pursue business opportunities.

For example, Deutsche Telekom, e&, Singtel and SK Telecom have formed an alliance to accelerate AI-first transformation of their businesses, and develop new, AI-powered business models.

AI factory

A big part of building an agile, product-centric tech-co is the ability to innovate at speed, and then innovate at scale using an AI factory. This method treats the creation of new AI IP as an industrialized process, akin to a factory production line, and emphasizes standardization, efficiency, and continuous improvement. Tech-cos should start by seeding the idea of a minimum viable product, define a measurement framework, evolve the AI foundation and responsible AI framework, and then use telemetry to evolve autonomous operations and measurement. Finally, the product is operationalized for use, and performance learnings are fed back into product design for the next iteration.

Transformation management

This foundational lever ensures the right change initiatives are in place, and they are going at the right speed, with metrics to indicate what’s working and what isn’t. Transformation management defines both a framework and roadmap for moving from a telco to an AI-first tech-co, taking into account the operating model itself, along with the main vision for the whole firm. It also ensures that capabilities are embedded within the organizational structure, processes, and culture.

We recommend taking a micro-change management approach; Infosys used this approach to transform from a legacy IT services company to a live enterprise within three years. By breaking off chunks of the transformation into bite-sized implementations, and scaling quickly, we saw exponential business benefit.

Moving ahead

A big part of this operating model, is again, the human-centric factor. All people in the organization must be on board the AI-first vision, and able to evangelize its benefits outside the organization. This builds a culture of human-machine collaboration that is greater than the sum of its parts and ensures that employees don’t take fright as the business transforms. As Jensen Huang, NVIDIA CEO, recently told Infosys: “While some worry that AI will take their jobs, someone who is an expert in AI will certainly do so.”

The time is now for telcos to build an AI-first operating model. The very best telcos, including SK Telecom, are using similar operating models to execute on B2B opportunities to build vertical businesses in healthcare and manufacturing, and to embed capabilities such as marketing, sales, software development, and even sustainability into their products and solutions. Those that follow suit will ensure their investments not only increase efficiencies, but create more opportunities.