At this year’s‘AI in Financial Services’ conference in London we hosted a pannel session with our partners at DNB. The session was titled ‘Revolutionizing Banking CX: Leveraging Conversational AI to Supercharge Second-Line Support at DNB’ and featured insights from Product Manager Maia Sognefest and AI Trainer Jørgen Hansen, both of DNB, as well as boost.ai’s CCO and Co-Founder Henry Vaage Iversen. Together, we were able to share their story of success with Juno, a virtual agent that has revolutionised internal support.
Meet Juno: Your knowledgeable virtual colleague
The session kicked off with Maia introducing us to Juno, the star of the show. Juno is an internal support virtual agent designed to make information readily accessible for DNB's customer service agents. Juno is a bit of a know-it-all, with in-depth knowledge about the bank's products, routines, and customer service guidelines, but it doesn’t hoard that knowledge; Juno shares it willingly and efficiently. Juno has quickly become a popular colleague at DNB, helping agents do their jobs more effectively.
The Journey to Success:
DNB's journey with Juno began with recognizing the inefficiencies in their previous system. Customer service agents relied on SharePoint sites, which contained extensive lists of internal routines but were cumbersome to navigate. The need for change led to the birth of Juno, the internal virtual agent. DNB had prior experience with virtual agents, having successfully launched an external bot with boost.ai called Aino in 2018, so Juno was a logical progression to a more streamlined customer experience.
So, what were the key ingredients for DNB's success in implementing Juno? Maia and Jørgen highlighted three crucial factors:
- Understanding users: The DNB team had firsthand experience as customer agents, which gives them a deep understanding of their users' needs. They spend a lot of time with users, leveraging the proximity of an internal virtual agent to build a solution tailored to their requirements.
- Focus on quality: DNB emphasised the importance of continually improving Juno's capabilities and the quality of its responses. Regular feedback from users helps refine Juno's performance, with around 500 feedback submissions processed every month.
- Autonomy and trust: DNB's team enjoys autonomy in managing Juno, aligning its goals with the organisation's objectives. This trust allows them to take full ownership of Juno's development and operation.
Balancing positive change and adoption:
Maia and Jørgen spoke about recognising that driving change in a large organisation can be challenging. However, they emphasised that trust and passion play a significant role in ensuring the positive outcomes they have seen with Juno. By involving early adopters and engaging users in the process, they created a sense of excitement and ownership among the staff that has continued to this day.
Harnessing generative AI and LLMs:
The discussion then shifted to the growing role of Generative AI and Large Language Models (LLMs) in boosting AI efficiency. While DNB acknowledged the potential benefits of these technologies, they also highlighted some risks, particularly in maintaining control over responses.
Two key challenges emerged:
- Variability in responses: Generative AI can provide varying responses to the same question, which may be desirable in some contexts but problematic for consistency in banking routines.
- Unstructured source data: DNB's source data sometimes lacks clarity, making it challenging for the virtual agent to provide complete and accurate answers.
Henry explained how boost.ai incorporated Generative AI into their virtual agents without sacrificing the accuracy of responsesvia a hybrid approach that leverages the company’s robust Natural Language Understandng (NLU) engine. DNB anticipated significant time savings in building and maintaining Juno's intelligence through features like training data generation, content suggestion, and improved language processing.
Overall, it was a highly engaging and enjoyable session covering how one of our close partners, DNB, has been able to support employees internally with conversational AI. To summarise, there were 5 key learnings from the session:
Key Takeaways:
- Team, culture, and trust: With the right team, culture, and trust, organisations can successfully implement transformative technologies like virtual agents.
- First impressions matter: A positive first impression is crucial when introducing organizational change. It can significantly influence the adoption of new technologies.
- Identify early adopters: Find innovators and early adopters within your organization who can drive excitement for new products and facilitate a smooth transition.
- Balancing control and innovation: While harnessing advanced AI technologies, it's essential to maintain control over responses, especially in highly regulated industries like banking.
- Continuous improvement: Commitment to quality and constant improvement is key to the ongoing success of virtual agents.
To find out more about how conversational AI agents could help to support your employees, get in touch with us today at www.boost.ai/contact
