If there’s one thing the world can’t seem to stop talking about it has to be generative artificial intelligence (GenAI). But even though it might seem like we’ve reached peak hype it’s important to differentiate between GenAI for consumers and GenAI for B2B.
“From a consumer perspective, it might seem like a hype,” says Aparna Khurjekar, chief revenue officer for business markets and SaaS (software as a service) at Verizon Business. “However, I’m excited about combining data and AI, as we've been involved with for a long time. Now, with GenAI, we can add creativity to our capabilities, creating a lot of potential.”
Don McGuire, chief marketing officer at Qualcomm, is equally excited. “GenAI is here to stay and moving fast,” he says. “It's crucial for us to embrace, learn, evolve, and figure out how to best use it.”
Leverage GenAI for new capabilities
Telecom operators play a dual role as providers and consumers of GenAI. This involves connectivity solutions, network planning, radio-frequency engineering, predictive maintenance, and more. By integrating GenAI into aspects like sales, intelligent meetings, and support, companies can provide personalised solutions for customers.
“The objectives revolve around three pillars for a mobile operator or any entity: enhancing team productivity, optimising the network, and developing customer-centric use cases,” says Eugina Jordan, chief marketing officer of the Telecom Infra Project. “For consumers and enterprise clients, it’s about leveraging GenAI for new capabilities, with mobile operators playing a pivotal enabling role.”
Leverage data previously held in silos
Although all this might seem brand new, AI has long been integrated into operations, such as determining the next best action for customers, product recommendations, and smart pricing strategies. Indeed, McGuire points out that Qualcomm has been working on AI for 15 years, with GenAI “bringing it into the foreground”. Khurjekar speaks similarly about its history at Verizon.
“GenAI’s advent introduces advanced applications like sales sequencing and personalised email generation in the field, and the deployment of info bots that navigate through retrieval and generation models,” she says. “This leverages data previously siloed within our databases.”
Keep up with rapid technological advancements
Still, even with all the potential, McGuire says that “about 80% of enterprises are not ready for generative AI, dealing with concerns and uncertainties, while 20% are moving forward”. But as CMO, he believes it’s crucial to efficiently use resources to keep up with rapid technological advancements. The same goes for employee productivity.
“We’re still early in this process but I don’t subscribe to the narrative that GenAI is going to replace jobs,” he says. “It’s meant to replace tasks. New skillsets have to be learned, whether it’s prompt engineering or learning how to use these tools. But it’s going to free up people to be better critical thinkers, to have work-life balance, and to create opportunities for career development or training… Our productivity gains early on are upwards of 20% to 30%.”
Khurjekar is seeing similar benefits from using GenAI to augment skills. “At Verizon, we focus on reducing the cognitive load for sales reps or individuals to allow them to handle more complex tasks,” she says. “We’ve launched a significant initiative with automation and AI, resulting in a 16% improvement in rep productivity. This increase enhances customer experience, generates more revenue, and benefits us.”
Create a standard industry model
How should businesses move forward with GenAI? One suggestion for the telco industry is to have a standard large language model (LLM). This would involve using a pre-trained model and fine-tuning it for the industry, providing a common baseline for model development. There might also be the need for either a highly curated small language model or a large action model focused on outcomes.
“The success of GenAI implementation hinges on two factors: the right workforce and access to data,” Jordan says. “A specialised LLM in telco is essential because regulated industries face unique challenges. Using a general model across these sectors could lead to significant issues.”
Use multiple approaches depending on need
Jordan also emphasises that the data should be open source to benefit the entire industry. At the same time, the industry must maintain its focus on privacy and security so as not to break customer trust. This leads to questions about where the data should live. “Does it stay on the device you’re using, on your private network where it’s protected and secure, or does it live in the public cloud?” Khurjekar asks. “There are arguments for all three depending on the data.”
Of course, the answer doesn’t have to be one or the other. “We’ve deployed two different chatbots,” McGuire says. “One is trained only on Qualcomm data, ensuring searches only access information within the allowed dataset. We also have a Microsoft Teams chat trained on public data. We deploy both private network-driven tools and public cloud-driven tools, so the choice depends on the context.”
Khurjekar is also seeing models emerge in hybrid network solutions. “With AI, there’s a lot happening with private, customised models at the edge, and several that need to connect back to the cloud or other databases,” she says. “Verizon is helping provide the right fabric and infrastructure for programmatic switches and routers for prioritised traffic management, especially in distributed private networks and edge compute solutions.”
Use the right kind of data
One criticism of GenAI (and AI in general) is that the tools can feel like black boxes that create answers without anyone understanding why. But explainability is important when it comes to getting full support.
“We have churn propensity models, and understanding why someone is in a high decile to churn versus not requires transparency and explainability, so you know exactly how to act,” Khurjekar says. “With Verizon, it’s a big opportunity. We’re setting the right kinds of guardrails and talking about governance models. It’s about what kind of data goes in and how we can start thinking about the outputs in conjunction with the bias that can show up. But it all starts with bringing in the right kind of data.”
Add the human element
Indeed, when it comes to accountability, it’s important to consider how the model was trained in the first place. For example, Qualcomm doesn’t deploy any tool unless the legal department is comfortable that the algorithms are both legal and that the data is either open source or licensed, for example Adobe Firefly licensing with Getty Images. Similarly, the focus on ensuring quality input also extends to ensuring quality output.
“You can’t just take that output verbatim and put it out to the world,” McGuire says. “You have to create accountability around it. In the creative process, these tools are great for ideation and for kick-starting the process, but you have to add the human element. You have to add that editorial aspect if it’s text-driven or copy-driven. You have to add the artistic element if it’s visual or from a video or photography perspective. If we don’t add that human element, we can’t own the content and the output. Our legal department has been very clear. So everyone has to go through training. That’s how we’ve deployed accountability and how we use these tools.”
Implementing GenAI in B2B
To leverage GenAI effectively in the B2B sector, Khurjekar suggests addressing specific business needs and achieving tangible ROI.
- Identify business needs
Begin by reversing the typical approach to technology adoption. Instead of starting with the technology (be it GenAI, deep learning, or machine learning), start with the core needs of your business. What challenges are you facing? What processes could be optimised? - Evaluate use cases
Once you identify needs, explore how GenAI can address these specific issues. Look for areas where it can have a direct impact on efficiency, cost reduction, or revenue generation. Examples in telco include network planning, customer service improvements, and operational optimisations. - Implement with purpose
Choose applications of GenAI that align with your strategic goals. In telco, this might mean employing it for geospatial intelligence in cell planning or for spectrum intelligence. The key is to focus on applications where it can provide a clear rationale and a solid return on investment. - Enhance customer service
Consider starting with customer service improvements as a practical entry point. Use GenAI for smart call routing, supporting agents with guided paths for sales or service, and integrating root cause analysis and next-best-action recommendations. These applications can quickly show value by improving customer satisfaction and efficiency. - Monitor and adjust
Continuously evaluate the impact of GenAI on your operations. Are you seeing the expected benefits? Adjust your approach as needed, always aligning with your core business needs and strategic objectives.
The Gartner hype cycle
The Gartner hype cycle is a graphical representation used to illustrate the maturity, adoption, and social application of specific technologies. Developed by Gartner, a leading research and advisory company, this model serves as a tool for understanding the progression of a technology from conception to mainstream adoption. The hype cycle is divided into five key phases:
- Technology trigger
The initial stage where a technology is conceptualised or introduced. Interest begins to grow, often without any practical products or proven applications. - Peak of inflated expectations
During this phase, early publicity generates a significant amount of hype around the technology, leading to inflated expectations. Success stories, often accompanied by scores of failures, start to emerge. - Trough of disillusionment
As experiments and implementations fail to deliver, interest wanes. Technologies that survive this phase continue to be improved and evolve. - Slope of enlightenment
More instances of how the technology can benefit the industry start to crystallise, leading to a better understanding of its practical applications. Early adopters begin to experience success. - Plateau of productivity
The technology reaches mainstream adoption, its real-world benefits are recognised and accepted, and the technology becomes increasingly stable.
Learn more: www.gartner.com
Gallery
