Generative AI Deployment Offers Telcos Significant Commercial Opportunities, but Traditional Shortcomings Means That They Are Unlikely To Seize Them

The public release of chatGPT in late 2022 has accelerated the democratization of generative AI and has spurred enterprise vertical interest in its commercial opportunity. Enterprises are looking for ways to utilize generative AI across business processes to lower costs and drive new revenue streams. The telecom sector is no different, as stakeholders across the value chain look unlock their commercial value.

Telcos have started implementing generative AI to improve low-risk business processes. LLMs are being used to support customer service by better understanding customer queries and generating responses. It has been implemented within AT&T’s Watson, Orange’s Djingo, Telstra’s Codi, and Vodafone’s Tobi to handle routine and repetitive troubleshooting tasks. By embedding generative AI into these chatbots, human agents are freed up to handle tougher, more time-consuming problems. This will lower operational costs and time-to-support, potentially improving customer retention in both the B2B and B2C markets.

Platform providers build out telco-optimized generative AI solutions. VMware and Wallaroo.ai have built a cloud platform to lower the barrier to deployment of machine learning at the edge of the network. Whilst, NVIDIA and Softbank have teamed up to build a platform capable of running telecom-ready generative AI services across distributed data centers in Japan.

Some telcos have ventured out alone to build out their generative AI services. SK Telecom are building an AI-based chatbot application, ‘A’, on-top of OpenAI’s GPT-3. The aim is to build a generative AI based ‘super app’ that integrates various services such as music streaming, e-commerce, and payment applications (which includes a variety of third-party services). China Telekom are looking to build generative AI services from the ground up with heavy investment in LLMs.

Although the telco sector has begun embracing generative AI, further revenue streams and cost efficiencies are certainly available:

• Build new revenue streams.
  • Utilize distributed architecture and build data center capacity to support enterprise edge generative AI deployment.
  • Build consumer facing applications and sell as value add-ons e.g., enabling real time voice translation to support communication between languages; real time voice-to-text to support accessibility.
• Support existing revenue streams
  • Develop hyper personalized services, for both B2B and B2C sales, by deploying LLMs to analyze customer requirements.
  • Improve customer experience by deploying consumer facing generative AI applications e.g., enabling real time voice translation to support communication between languages.
  • Improve customer service by deploying intelligent assistants/prompts to guide conversations.
  • Support sales representatives by automating pricing for B2B and B2C engagements.
• Cut costs across business processes.
  • Reduce customer service burden by implementing LLM-based chatbots for repetitive troubleshooting jobs, whilst retaining human support for more challenging issues.
  • Lower network developer costs by using generative AI to create software code.
  • Perform proactive network maintenance by assessing data trends with LLMs.
  • Reduce time and cost of infield network maintenance by providing intelligent assistants.

Telcos have increasingly used artificial intelligence and automation across their network, now they must look at ways to embed generative AI across business processes to develop new revenue opportunities and help cut costs. But how can telcos get the most out of generative AI?

Partnering is key. Although some telcos from Asia Pacific are looking to build generative AI services from the ground up, this would be a risky strategy to follow as most tier 1 or 2 telcos lack the hardware/software skills necessary to ‘go-it-alone’. Partnering will be quicker, easier, and lower costs allowing operators to more efficiently access the commercial opportunities associated with generative AI.

But who should telcos partner with? Generative AI has a long value chain with three partner options – foundation model providers (like AWS, OpenAI), application developers/ISVs, and enterprise service providers. Each has a role to play, but given the scope of transformation required within telcos business processes, operations, and human capital, turning towards the enterprise service provider will be most effective, as they can help:

• Identify opportunities and align deployment with telco capabilities, priorities, budget, and strategy.
• Provide end-to-end generative AI solution with support across day 0 and 1 operations.
• Build customized models within telco walled garden.
• Choose the right LLMs (i.e., open vs closed sourced, model parameters, on-premises vs cloud deployment) and fine tune to support telco specific use cases.
• Implement and customize governance and guardrails (security features) to manage employee usage and data security concerns.
• Deploy operational changes and train employees to support the process of adopting generative AI.
• Provide support for day 2 operations (this is vital given that LLM implementation should follow an iterative approach to optimize performance, cost etc.)

The immediate focus for telcos should be to deploy low risk use cases quickly (i.e., B2B sales research and market trend analysis, intelligent assistant for customer service representatives.) Of course, given the data privacy challenges associated with generative AI, telcos must be cautious; but its is vital that they avoid protracted PoCs as building competitive differentiation requires quick deployment across business processes. The best and fastest route to deployment is to partner with either system integrators or business consultants as telcos neither have the technical nor transformational skillset to overhaul business processes through generative AI deployment.

But do we really expect telcos, that have struggled with transformation initiatives, to successfully seize new commercial opportunities? No. Certainly, we see implementation in isolated use cases, perhaps customer service or in-field intelligent assistant, contributing to some cost efficiencies. But we are highly doubtful that telcos will be able to drive new revenue streams; this is not down to the technology, but more the traditional transformational shortcomings of the telco.