Six opportunities for ISPs to Leverage AI
Jason Presement
Business Development and Sales Leader ? Technology Evangelist | Growth & Innovation | Trusted Advisor | Strategist | Industry Relations | Inspirational Leader | Newsletter Publisher | Telco, Space, AI
If the DALL-E generated image above is any indication, AI still has a ways to go towards instilling confidence in users that the decisions being made and, in this case, the output being provided are correct and demonstrate the necessary and intended "intent."
This becomes critical as we move past Generative AI tools like ChatGPT, Google Gemini and Perplexity to fully intent-based, autonomous tools that could be leveraged by ISPs to enhance their business - be it focused on boosting efficiency, reducing cost or affecting an internal user or support experience or an external subscriber experience of some sort.
As I participated as a panellist in a Fiber Broadband Association webinar this past week that focused on how ISPs can leverage "AI" in various ways, I drew upon my own conversations that I have had with many ISP as well as some anecdotal findings and reports from real-world implementations to summarize the information as supporting materials for the webinar.
With that, in no specific order, here are the six areas that I believe provide opportunities for ISPs to leverage the power of "AI," whether based on machine learning, large language models or more advanced models, to positively impact internal and external experiences across the organization. This is by no means an exhaustive list, but rather meant to foster discussion and thought.
1: Customer Support
Chatbots are by far the leading application being affected by AI tools and technologies. A recent report from AWS and Altman Solon suggested that in their survey, over 63% of the respondents had some sort of chatbot application in production. I see this as the low-hanging fruit when it comes to affecting outcomes and experiences for ISPs.
In the realm of customer support, the report further breaks down three opportunities to leverage AI in customer support applications beyond "simply" having a natural voice chatbot interact with inbound customers. Although all of them rank highly, clearly, chatbots remain top of mind:
At a high-level, the concept with chatbots is simple; a human, conversational voice answers the inbound call, triages the issue and provides a fix for the caller where possible.
For example, a user calls in to the ISP because they can't remember their WiFi password. They have been validated and verified by their voice-print, and perhaps another challenge question. The caller tells the chatbot, in plain language, that they forgot their WiFi password. The chatbot looks up the information, provides it to the caller and perhaps sends a follow-up email. The customer is happy. An agent wasn't required. First Call Resolution and call time are low. Everyone is happy.
Another example may be more focused on an issue with a response that is more "learned" rather than one that can be looked up as a parameter. For instance, I used to hear a lot about customer support people who would have to walk more senior users through the process of shutting off, or turning on, the Secondary Audio Program (the text in the TV) that may have accidentally been switched on, or off. A conversation with the chatbot around "words on the screen" and whether they wanted it or not would lead to a quick, non-frustrated resolution, again, without impacting client representatives who may be involved in more complex problems.
Of course, this is an example of where support models need to be trained using thousands of hours of existing support recordings or possible community-sourced information from a provider collective to cross-train models.
Some other advantages include the ability to answer a call immediately and not put a caller on hold, to deal with a diversity of languages, to triage problems more effectively and to offload customer support representatives to focus more on complex problems.
Of course, this should ultimately positively affect the ISPs operational metrics, KPIs and OKRs, including NPS and overall satisfaction.
Two other internally focused support applications also need mention.
There are examples of LLMs being used to help the customer support representative more clearly both define a user problem, and also search a knowledge base on a multiple of symptoms or problems linked to an issue towards a better overall resolution. In addition, AI-based oversight of the customer service representative, providing suggestions on what, how and where to query based on the problem and progress of the troubleshooting to date, both trains the model further and provides for quicker resolution - all through natural language interactions.
Finally, analytics and insights collected to inform the need for further agent coaching and other related activities.
2: Fraud Detection and Security
There are many examples of where AI-based tools could be used to mitigate fraudulent activities. Whether this a priority for ISPs to address in the face of other opportunities to affect other customer experiences will become clear over time.
Areas to impact fraud are as follows;
3: Predictive Maintenance
One of the key advantages of AI-driven predictive maintenance is the ability to anticipate potential issues before they cause service disruptions.
For example, AI systems can analyze patterns in network traffic, equipment performance data, and environmental factors to identify early warning signs of impending failures. This allows ISP technicians to address problems preemptively, reducing downtime and improving overall service reliability for customers.
AI tools can also help ISPs optimize their maintenance schedules and resource allocation. Instead of performing routine checks on a fixed timetable, predictive models can determine when specific components or sections of the network actually need attention.
Another interesting application is in the realm of capacity planning. AI-powered predictive maintenance can forecast future network demands based on historical data and emerging trends. This allows ISPs to proactively upgrade infrastructure and allocate bandwidth or scale in areas more efficiently, ensuring smooth service even during peak usage periods.
The impact of AI on predictive maintenance goes beyond just fixing things before they break. It can also enhance the customer experience by minimizing service interruptions and providing more accurate estimates for resolution times when issues do occur.
As discussed earlier, ISPs can explore using AI chatbots and virtual assistants to handle basic troubleshooting, freeing technicians to focus on more complex problems.
While the potential benefits are significant, implementing AI-based predictive maintenance does come with challenges. ISPs must invest in robust data collection systems, train their staff on new technologies, and carefully manage the transition to avoid disruptions.
Having said that, as AI tools become more sophisticated and accessible, the opportunities for ISPs to leverage predictive maintenance will only continue to grow, potentially transforming the industry's approach to network management and customer service.
4: Network Optimization
The engineering folks like to start here. In fact, during the webinar one attendee suggested that if you solve for NetOps first, everything else on the support side falls in line. I see it as more part of the solution, although having NeOps in place does open up additional automated customer support actions. However, you need the customer support side in place to take advantage of the overall ecosystem. Chicken and egg. Cart and horse.
领英推荐
Either way, ISPs are increasingly looking to understand how can assist with network optimization, addressing the growing complexity of modern networks and meeting the rising customer expectations. Many ISPs have automation in place today, but do they have true end-end AI across multiple vendors and domains? Not yet
Examples of where AI NetOps add possible value include;
Traffic Management: AI can optimize network traffic by dynamically adjusting bandwidth allocation based on usage patterns and demand. Machine learning models can analyze historical data and real-time traffic to predict congestion and automatically reroute data to less crowded paths. This ensures smoother performance during peak usage times and improves overall network efficiency. This can be both inter- and intra-domain based, self-healing, intent-based, and all that
Automated Configuration: AI-powered systems can learn optimal network configurations and automate the process of adjusting settings. This reduces human error in network management and allows faster adaptation to changing network conditions. As networks become more complex, especially with the rollout of 5G, AI can handle the intricate task of balancing multiple network parameters for optimal performance.
While we're talking about 5G, AI-driven network optimization can significantly enhance the efficiency and effectiveness of network slicing, which allows a single physical 5G network to be divided into multiple virtual networks, each tailored to specific use cases with distinct performance requirements.
For instance, AI algorithms can analyze vast amounts of real-time data to predict network congestion and dynamically allocate resources to different slices based on current demand and performance needs. This ensures that critical applications receive the necessary bandwidth and low latency while less critical services are managed appropriately to maintain overall network performance.
AI can also help design and adjust network slices in real-time by continuously monitoring network conditions and user behaviours. AI can optimize the configuration of each slice to meet changing requirements. For example, during a major live event, AI can predict increased data traffic and adjust the network slices to prioritize video streaming, ensuring a better experience for viewers.
Customer Experience Enhancement: AI can help ISPs offer personalized services and proactively address customer issues by analyzing user behaviour and network performance data. As we discussed earlier, natural language processing and AI chatbots can handle basic customer queries, while more complex issues can be routed to agents with AI-assisted diagnostics, improving response times and customer satisfaction.
Of course, there are also some associated challenges. These include the high initial costs of AI systems, the need for specialized talent to develop and maintain AI solutions, the complexity of integrating AI with existing network infrastructure and working across multi-vendor environments with multiple and diverse data sets. There are concerns about data privacy and security when using AI to analyze network traffic.
Some wonder if the expense yields real value, whether OPEX-related or affecting overall NPS or customer experience. Is the juice worth the squeeze? That's the NetOps AI question,
Either way, AI in network optimization is compelling. Networks certainly aren't becoming "more simple" and less important. AI will likely become an essential tool for ISPs to maintain a competitive edge, improve service quality, and manage costs effectively. The key will be to focus on areas where it can provide the most immediate value while building the foundation for more advanced applications in the future.
5: Revenue Growth
AI offers ISPs a wealth of opportunities to enhance their service offerings, create more meaningful and profitable customer interactions, and drive enhanced ARPU through various strategies, including upselling, cross-selling, personalized product recommendations, and new product development.
Upselling and Cross-Selling: AI can identify patterns and preferences by analyzing customer data, allowing ISPs to make targeted recommendations. For instance, if a customer frequently streams high-definition content, AI can suggest upgrading to a higher bandwidth plan. Similarly, AI might recommend adding a streaming service or a smart home security package if a customer has a basic internet package. This personalized approach increases the likelihood of a sale and enhances customer satisfaction by offering relevant and valuable services.
Personalized Product Recommendations: By analyzing user behaviours and service usage, AI can tailor product recommendations to individual customers. For example, if a customer frequently uses online gaming services, AI tools can recommend a gaming-optimized internet plan with lower latency and higher speeds. This level of personalization can lead to higher conversion rates and increased customer loyalty, as customers feel understood and perhaps valued.
New Product Development: AI can also play a pivotal role in new product development for ISPs. By analyzing market trends, customer feedback, and competitive offerings, AI can identify gaps in the market and suggest new services or features. For example, AI might identify a growing demand for cybersecurity services and prompt the ISP to develop a new product line focused on protecting customers' online activities, complete with a business plan and business case, including go-to-market plans, revenue forecasts and suggestions on targeting customers for the optimal market penetration and success.
Revenue Growth: According to a report by McKinsey, companies that adopt AI in their business strategy have the potential to increase their revenue by up to 20%. This is because AI helps identify new revenue streams and optimizes existing ones by making operations more efficient, personalized, and customer-centric.
As we've seen with other areas of AI, there's an investment required to support a robust and secure data environment, as well as for internal resources to build, scale and manage the supporting systems. The increased revenue and reduction in churn play favourably in the business case.
As AI technology evolves, its role in transforming the revenue side of the ISP business will only become more pronounced.
6: Sustainability
Last, but not least, is sustainability.
Intelligent power management in fixed and wireless networks is becoming increasingly important for ISPs and telecom operators. With the rising costs of electricity in some markets to support the expansion of network and support systems, saving a few KWh here and there can make a significant difference.
A couple of examples of ways in which AI can be used to not only meet ESG goals and KPIs, but also save some money are as follows.
Anything that reduces power, optimizes cooling, reuses generated heat, and so on will further sustainability developments. I have no idea what the carbon footprint of the systems required to develop sustainable solutions is.
In summary, net of GenerativeAI, which is becoming fairly well understood—or rather being used by more and more people every day—we're still in the early days of things. Many people look to real-life examples to understand actual business cases and results—to either support an investment to grow or maintain the status quo.
We'll see how competitive pressures prioritize solution adoption within the ISP community and whether solution adoption and prioritization are influenced by the size of the subscriber base.
AI is coming fast, and people are looking to justify the spend. We need more case studies and real use cases to help make that happen.
What do you think? What are you seeing? What have you deployed and what results have you seen?
By the way, I have a weekly newsletter called "Jason's Industry Insights" that focuses on broadband, space and AI. To have it delivered to your inbox weekly, click HERE!