Many organizations are eager to harness the promise of AI to streamline operations, particularly in areas such as support and helpdesk. But Jana Bohackova, IT Store Project Manager at electronics retailer Elkjøp Nordic, cautions that AI will not deliver results if your data is not up to the mark. She warns:
AI will not solve anything if your data is wrong. You need to ask yourself, are you actually ready to use AI? Of course, it's amazing. It will change everything. But you need to think about the data.
Elkjøp is a long-established electrical retailer with a strong presence in Norway, Sweden, Denmark and Finland. Part of the UK-based Currys retail group, it has around 400 stores across the Nordics and Iceland, as well as a store in Ireland.
An IT team of about 100 employees supports a massive catalog of tech equipment in the stores, including over 80,000 PCs, 3000 PDAs, and over 1 million electronic shelf labels. In addition it has over 250 applications, and over 250 microservices to service.
It has been using Atlassian's Jira tools as a core part of its IT and support organization since 2008, with Jira Service Management acting as a central platform for tracking and managing issues across departments and locations. But the sheer volume of tickets was generating a growing support headache. Bohackova says:
Support was becoming more difficult, with the help desk a central problem. The company had 80,000 support tickets in service management raised in the first 8 months of the year, and that's 62,000 support tickets raised by the users in helpdesk -- around 350 things on a daily basis, and we only have 100 IT people for the core operation. We had 62,000 reasons to look at optimization, automation and AI.
The team began a project to harness AI earlier this year, by introducing Atlassan's Virtual Service Agent for self-service queries and ticket creation. This is an AI agent that uses natural language processing and generative AI to give conversational support to customers. It allows users to ask more detailed questions, and will understand the context of the conversation. But the initial deployment wasn't all plain sailing. Bohackova continues:
We first heard about the Virtual Service Agent last October, and we were eager to get started. We wanted to enable our team and give them time to focus on more complex work, not just answering questions. So we started using Atlassian Intelligence Answers, which is a part of the Virtual Service Agent.
The project framework was established in May. Then we turned it on and realized it was wrong. We looked at the answers, and we found that our data quality was just not good enough, so we had to go back and try again.
The company had to return to its knowledge base, which is within Atlassian's Confluence. It used Confluence AI to provide better input for its support data. Its goal was to achieve a 15% reduction in ticket integration over the next six months. This works out at a reduction of 40,00 hours over the year, says Bohackova. Even though the rollout won't finish until next week, it is already making an impact.
She says that the agent will allow Elkjøp to improve and increase its support, without having to increase the number of staff. Another benefit is that it is available 24/7, even at the weekends when retailers are at their busiest.
Bohackova advises considering what the challenge looks like from the beginning, what you need to provide, and what you need to build. She says:
It's not required to have a special skill set, not really. You just need to understand the process. You need to understand how the user thinks, and what you would ask if you were in the place of the person.
That's important because sometimes people forget that they are too technical. So you sometimes need to look at it in the real user case and talk to people, what do you trust to get the best answer on how we will build the content.
Bohackova was speaking at last week's Team '24 Europe conference in Barcelona. During her session, a number of new features for the Virtual Service Agent were showcased, which will be available in the next few weeks. These include multi-language support, of particular interest to Elkjøp. It allows users to ask questions in their language, and the AI provides answers based on existing knowledge articles in English. This removes the need to translate knowledge, so whether a question is asked in Spanish or English the answer will be available in the questioner's own language.
Another feature that will soon be available is the Virtual Service Agent on Email. This will address close to 50% of inquiries that still arrive at the Elkjøp helpdesk through email, providing answers to user queries with links to relevant articles. It will also allow users to give feedback and agents will be able to track which answers were helpful.