Banking services forerunner POP Bank harnesses AI for its contact center
Based on feedback from business and consumer customers, POP Bank has been chosen as the best bank in Finland for 13 consecutive years. “We offer centralized contact center service to our alliance banks,” explains Pia Ali-Tolppa, CEO of POP Bank Group’s internal service provider, Bonum Bank. Then she modestly continues: “We naturally cannot take all the credit for this accomplishment, as our 18 independent alliance banks meet the customers and have a tremendous impact on this result, too. But surely, partially also, the service our contact center people offer is the reason for our delighted clientele.”
What was the challenge?
Despite her current business executive role as the CEO of Bonum Bank, Pia Ali-Tolppa has a keen interest in technology’s effect on people’s work. “I have a degree in programming but have never been working in that field and do not consider myself a technology enthusiast,” she conveys. Nevertheless, Ali-Tolppa’s passions are artificial intelligence and its potential to make people’s everyday lives more efficient and meaningful.
Thus, it was no surprise that when large language models (LLM) like Chat GPT were launched to the public in the turn of 2023, Ali-Tolppa was among the first ones to try it in her own everyday life and started wholeheartedly talking about generative AI’s value for financial services at work, too. Soon, she heard from their IT department that POP Bank’s cloud platform partner, Google, offers support to customers’ AI and LLM pilots that run on Google’s Vertex AI platform.
Finding the correct information is the most stress-inducing part of the agents’ work.
“Our agents are drowning in documents and instructions,” she explains. “It is impossible to memorize everything, which leads to cumbersome detective work and slows down customer interaction. When the agent needs to pick and choose pieces of information from different data sources, errors, too, become more common.” Ali-Tolppa points out.
Bonum Bank’s internal employee satisfaction surveys also revealed that finding the correct information and answers for a customer was the most stress-inducing part of the agents’ work. This was when Bonum Bank got the idea of harnessing AI to help the agents.
Our solution
“The thinking from the beginning was to augment the human – not to replace them,” says Ali-Tolppa. Bonum Bank did not want the customer to interact with the language model but instead wanted the agent to serve between the AI and the end customer. When a customer calls, the generative AI chatbot finds and summarizes the relevant documentation based on a natural language query made by the agent.
“To me, this seemed like the perfect pilot,” rejoices Ali-Tolppa and continues: “It is straightforward and does not require much development work or budget. And if we get good results, it is easy to expand to other areas.” Ali-Tolppa also stresses that when successful, this pilot would allow them to provide instantaneous relief to a significant burden to their contact center agents.
Why Codento?
After discussing their idea with Google’s team, Google introduced Bonum Bank to Codento. “Google complimented Codento for their long-lasting partnership and good results,” Ali-Tolppa explains. The pilot itself was implemented surprisingly fast, being almost a textbook example of agilely testing new technology and being ready to succeed fast. “I think it wasn’t more than a month from when we explained the need to Codento to when the first pilot was up and running for our test group of agents,” Ali-Tolppa says. The collaboration, overall, has been fruitful and smooth.
What were the results?
The results are promising now that the solution has been used in the pilot group for some months. “Our agent’s work has become substantially faster,” says Pia Ali-Tolppa. “Also, when new people come in, the solution gives them a great backbone to do their jobs and get on top of things quicker,” she continues.
Work has become substantially faster.
New working methods include learning, too. As in many conversational AI cases, Pia Ali-Tolppa mentions their challenges with correct prompting. Asking the right questions from the AI bot is crucial to get accurate and exhaustive answers. “Also, this way of working demands methods to keep the data up-to-date and enrich it. We need to discontinue storing documents and instructions in people’s folders. When we notice something missing, there must be a process for complementing the common databases, making data accessible to everybody,” she says.
With the excellent results of speeding up the strain-causing information searching and making onboarding new agents more efficient, Pia Ali-Tolppa envisions spreading AI usage in Bonum Bank and possibly the entire alliance of banks. “First, we will surely expand the use of AI to the entire contact center and all of our documentation, but I am sure that from these encouraging experiences, entirely new ideas will also come up,” she concludes.
| Google technology | Usage in POP Bank’s case |
| Vertex AI | Google’s platform for data science projects |
| Gemini LLM
|
Family of multimodal large language models (LLMs) used for generating the answers for the questions asked by customer support agents |
| Cloud SQL | Storing document embeddings and metadata |
| Cloud Storage | Storing raw documents |
| PubSub | Updating the vector database when data is added or removed |
| Cloud Run | Frontend and api hosting |
| Artifact Storage | Storing docker containers |
Contact us to learn more about our services:
