Skip to Content

Generative AI is making life easier for product support engineers

Nikhil Gulati & Jalaj Pateria
May 21, 2024
capgemini-engineering

Learn how Generative AI (GenAI) is revolutionizing Software Product Support and how to get started with this powerful technology in your business.

Generative AI (GenAI) is beginning to transform many activities, and product support is no exception. Product support is vital for the ongoing function of all products, from Microsoft Office to niche robotics systems. Users need product support when installing systems, integrating with other software, working out how to use the product, and resolving issues when they arise. 

Such work must be handled by experts who understand the product and its operation. The cost of this support must be factored into any product cost model, so improving the support process can unlock revenue by extending the life of products while reducing the costs of supporting them. This is particularly true as products reach “end of life”, when user numbers often shrink, and support costs relative to revenue can become problematic.

The potential of GenAI in product support

Because GenAI can process information and predict the answer to a question based on experience, it opens a world of possibilities for product support. Given sufficiently large training data of good quality, GenAI can be taught about the fundamental nature of systems and predict the most appropriate answers to questions about them. A few examples of GenAI’s potential uses in product support are developed below.

  • Tech support automation: GenAI’s ability to learn answers to common technical questions about problems and provide quick and detailed responses means such a service can be available 24/7. Further, GenAI responses can be adapted to the specific user query and context. This approach is an important improvement on the typical support model, based on asking a series of fixed questions and pointing the user to an off-the-shelf ‘how-to’ article.
  • Augmenting human support workers: GenAI can facilitate the work of human support workers by summarizing requests and providing these workers with the relevant information to solve these requests quickly. If support workers respond by email, GenAI can help them turn their response into text that will be easier for the user to follow, based on the GenAI model’s technical knowledge. It can also translate responses, allowing teams to offer support, even when they do not speak the user’s language.
  • Onboarding new hires in the support team: A support GenAI can be used to train new support engineers on common product issues.
  • Software product upgrades: Generative AI can be used by support engineers to facilitate software product upgrades, for example, translating software code into a newer language or modifying code to be more efficient as part of a green code sustainability initiative.
  • Streamlining processes: GenAI tools can automatically categorize emails and support tickets and learn to prioritize in order of importance, assigning these to the relevant experts or those with the most capacity.

A well-composed suite of GenAI-powered tools can reduce time-to-solution, human error, and product support costs and so allow experts to focus on the more complex tasks that humans are best suited to.  

GenAI in product support – the art of the possible

Theoretical possibilities are all well and good, but what is happening in the real world? Capgemini is fortunate to have worked with multiple clients on projects to create value by harnessing GenAI in their product support processes and systems.

In one example, a large computer hardware organization wanted a system to identify multiple ticket types, handle initial conversations with users, and respond in various languages. The GenAI system we developed provided the firm’s customers with step-by-step instructions on how to resolve their queries. These responses were based on information in product knowledge bases and user manuals. It also identified user queries that couldn’t be solved using this approach and then escalated them to human support engineers. Finally, the GenAI collated user feedback and used this to propose updates to the knowledge base. The outcome was considerably fewer tickets routed to human agents, saving time and money.

In another case, we worked with a Network Equipment Provider to develop a chat assistant to provide ‘human-like’ first-level responses and summarize tickets for efficient handover to other support staff. Again, we saw reduced operational costs and improved SLA (Service Level Agreements) adherence in their 24/7 operations.

In a final example, we built a do-it-yourself (DIY) tool and analytics generator for a leading telco. They needed to document the standard operating procedures (SOPs) of their support engineers for future training and generate role-based visualization and prediction. The customer required a centralized management dashboard that unified all IT platforms on a single pane and a GenAI-based tech stack for predictive and preventive monitoring. 

The challenges of integrating GenAI in product support

Developing, deploying, and running GenAI-powered systems is becoming ever more accessible, thanks to the increasing availability of large open-source language models. However, care needs to be taken when integrating AI into systems.

Firstly, GenAI must be carefully crafted and trained for the specific use case – using up-to-date, high-quality data. The AI will be wrong if the user manual or knowledge base is wrong. This means that people who understand the product for which the GenAI support system is being developed must be involved in designing and testing it. They must ensure it has been trained correctly. Because GenAI is probabilistic, GenAI outputs can occasionally be wrong; this is often described as a ‘hallucination’ in the GenAI community. Consequently, quality control is vital.

Secondly, there are IT practicalities to consider. The IT infrastructure must offer sufficient computational power to run a GenAI model and provide the connectivity needed for the GenAI to interact with knowledge management databases and issue management systems (including email, WhatsApp, etc.). There must also be a single source of truth so that any updates to the knowledge base – by humans or AI – feed into the GenAI’s model of the world. Organizations must be willing to share this data, regardless of its sensitivity.

Finally, GenAI project timescales need to be calibrated to the business case. Training takes time, but no business wants to wait a year for a perfect GenAI support system that will be obsolete when launched. An AI that can solve 50% of queries and refer the rest to humans but takes three months to build and deploy may offer better value than one that can solve 60% of queries but takes two years to deliver.

Ultimately, the recipe for success with support systems is the same as most data projects. Set clear goals and expectations. Work with experts who know the tech and the domain, and use frameworks that allow you to move efficiently through the development process.

Capgemini has multiple software frameworks and project blueprints to accelerate the development, deployment, and operation of GenAI in product support systems. Contact our experts to learn more.

Meet our experts

Nikhil Gulati

Head of Intelligent support and services
Nikhil is a results-oriented professional with extensive experience in IT/Telecom, Project Management, Software Development/support, Client Rela-tionship Management, Business development and operations, and Pre-Sales.

    Jalaj Pateria

    Enterprise Architect
    Jalaj is a Chief Automation Architect at Capgemini, Intelligent Support Services. He has over 16 years of experience working extensively on Digital Trans-formation Initiatives across BFSI, Health Care, Airlines, Industrial, and Telecoms. Currently working on next-gen initiatives in consulting, pre-sales, and solution phases, Jalaj’s research interests lie in Machine Learning, Explainable AI (XAI), Deep Learning, Sentiment Analysis, Digital Twins, AR/VR, and Automated Reason-ing.