Communications service providers (CSPs) are cautious about the GenAI revolution. For years they have invested significantly in AI, hoping to drive automation and efficiency across their business operations. While some initiatives have paid off, many have not yielded the anticipated revenue boosts and efficiency gains. CSPs are therefore more focused than ever on cost-benefit analysis. Industry leaders wonder: What will be different this time? If CSPs are going to invest heavily in GenAI, the business case needs to be profound.
Fortunately, it is. GenAI is poised to supercharge the outcomes of previous AI investments while opening potentially transformative new use cases, enabling telecoms to achieve substantially greater ROI than ever before. This will require CSPs to consider how GenAI apps, tools, models, and infrastructure will play out in terms of business outcomes, deployment time, and customer impact, and how GenAI can augment their current and future AI use cases.
Revisiting Past AI Use Cases
CSPs will need to revisit previous AI use cases and ask: Can GenAI transform the AI initiatives that have so far delivered only incremental success? In many cases, GenAI will increase the speed and accuracy of processes that have already been employing AI, finally allowing those projects to deliver the sought-after return on investment.
On the sales front, AI is already being used to offer personalized product information. GenAI will enhance guided customer purchases with an automated yet highly interactive experience. Similarly, when it comes to customer service, GenAI will optimize the performance of the virtual assistants that are already responding to tech troubleshooting and billing inquiries. In the cases of media services, advertising, and demos, AI already plays a role in the data analysis that optimizes image selection. With GenAI, the technology will create the images rather than simply identify and select them.
Data Mapping and Transformation
While the initial buzz around GenAI has been driven by its remarkable text-to-text and text-to-image capabilities, enterprises now recognize that its impact will extend far beyond chat text and jpegs. GenAI will also design and orchestrate systems.
Data mapping and data transformation is the most complex activity in BSS/OSS transformation for CSPs. GenAI will add agile, contextual nuance to data migrations, improve data rationalization, and perform real-time data translations between platforms for MACD types of transactions and new customer/order transactions. Further, GenAI will facilitate platform-to-platform integration by actively designing new business process flows.
Reimagining the Software Development Lifecycle
CSPs operate in a software-intensive industry, so it should be no surprise that telecom leaders are especially interested in GenAI’s code-writing abilities. If the software development cycle could be largely automated, CSPs would be able to reduce both service delivery cost and the product development lifecycle. They would no longer face a constant battle to attract thousands of developers who understand Python, Java, etc. Plus, the entire DevOps prioritization model (P1, P2, P3) could be retired, replaced by a system that delivers the needed code in seconds or minutes regardless of priority level.
In theory, automating software development sounds truly transformational. In practice, it will require CSPs to build a new function dedicated to monitoring these machine-learning initiatives. Even if most code generation (and even QA and code testing/ticketing) is automated, CSPs will need to spin-up talented diagnostic boards to provide human oversight of any code that is failed by exception, and retrain the AI to continuously improve its performance.
Network Design, Optimization, and Management
Network design is one area where GenAI could validate its return on investment rather quicky. GenAI is well positioned to play a role in redesigning the topology of networks by considering factors like capacity planning, required nodes, and utilization. CSPs could easily compare GenAI results to what they are getting from offshore/onshore network designers, then prove out and grow toward a full AI network design model in as quickly as 6-9 months.
Beyond the design phase, GenAI will also be able to improve service performance by contributing to real-time network optimization. Service performance can be close-looped with network planning tools as a feed to provide the most optimal performance. Also, given that networks will mostly be open moving forward, capacity allocation can be programed dynamically using GenAI to enable an advanced “network on demand” capability. This will be particularly impactful when it comes to event management: monitoring and auto-managing high profile occasions and sporting events.
The Right Lens for GenAI: Costs, Revenue, and Customer Experience
Given the industry’s broad AI track record, CSPs’ business and IT leaders will have little patience for large GenAI initiatives that sound interesting but will not prove their value in the near term.
One of the reasons CSPs have struggled to maximize the ROI of AI comes down to simple computing costs. Machine-learning algorithms are expensive to run, and the large language models (LLMs) that drive GenAI tools and infrastructure (including compute, network, and storage) will require even more compute than previous AI tools. Here, CSPs need to look beyond the hype and ask: Which functions truly need to be run on LLMs?
Even if GenAI tools can be deployed for numerous use cases, they should only be applied if they turn out to be more cost-effective than legacy tools. Leaders will also need to ensure that the anticipated impact and scale of each GenAI intervention matches the choice of GenAI model, whether that be platform-based private AI models (most expensive), commercial platforms like GPT-4 or Google BARD (less expensive), or customizable open-source models like Hugging Face (least expensive).
As business leaders begin to evaluate whether emerging GenAI initiatives will justify their costs, they can look to customer experience metrics like CSAT and NPS to validate the impact of GenAI on customer service. More broadly, they should look at process cycle efficiencies across all products. Because GenAI represents a more advanced level of automation, successful implementations should measurably increase service delivery speeds. Increasing service delivery speeds, in turn, will show up as accelerated revenue recognition, meaning that successful GenAI implementation will actively generate cashflow.
As they initiate and measure their first GenAI projects, CSPs do not need to achieve “big bang” transformations. Rather, they should pursue targeted proof-of-concept projects and build in mechanisms to quickly measure the bottom line, topline, and/or customer impacts. When the proofs-of-concept justify their expenses (in terms of both raw computing costs and monitoring/training), those successful Gen AI initiatives can then be scaled across the organization.