In the UK, Tesco employs more than 367,000 people. The Product Learning and Development (L&D) team at Tesco support thousands of these employees each day, providing them with training materials and guidance.
Due to legacy systems, colleagues were struggling to easily find the right information and with stretched resources, the L&D team were unable to provide 1:1 coaching on the learning options available for all their colleagues, so they needed to find a low-cost, high-value solution to better support their colleague’s needs.
We worked with Tesco to design and develop a chatbot solution known as Tess, that would be installed on their intranet to provide colleagues with rapid, reliable answers to their questions and guide them to the right online resource or in some cases training course.
Implementation
Tesco wanted colleagues to see Tess as an extension of the support they already receive from the L&D team, so when designing Tess’ we wanted the avatar and tone of voice to be friendly, approachable and helpful, mirroring the team’s communication style. At the start of every conversation, Tess introduces itself as a ‘learning assistant’ that was built to help answer ‘your’ questions. The tone is welcoming and personable, and the message clearly identifies what the service is for, why it was built and how colleagues can use it.
One of the main challenges colleagues faced when looking for information was identifying where to look and often the difference between legacy and new systems was unclear. Tess’ welcome message offers colleagues a clear place to start using buttons. We reviewed the team’s existing inbounds to identify common areas where colleagues request support and therefore where Tess could add the most value. These areas were grouped into four main topics which Tess presents to users as buttons in the welcome message. We then built conversation flows that help colleagues navigate these topics, to point them towards the right online resource.
As well as guiding colleagues using buttons, it was very important that Tess could answer questions via free text. Tess uses Google Cloud’s Dialogflow for Natural Language Processing and was trained to answer hundreds of FAQs, as well as respond to small talk. Tesco recognised that encouraging colleagues to self-serve and use a new channel for support may have been challenging. To raise awareness and support for Tess even before it was launched, the L&D team organised a poll asking colleagues to name the chatbot based on a brief description. The response from colleagues was very positive and creative. Introducing colleagues to Tess early on helped set expectations as to what the service was for and how they can benefit from it before they even opened Tess.
We could not and did not want to train Tess on the whole Learning and Development knowledge-based. The intranet where Tess is installed contains well over 800 pages and all this content would make Tess harder to manage. For the initial launch, we trained Tess to answer the most popular queries with a view to building its domain knowledge over time. We designed Tess to fail gracefully when it couldn’t answer a question. It would explain that it is still learning and will aim to have an answer next time, then give the colleague an alternative support route. We trained the Tesco team to use our EBM platform to monitor these misunderstood queries and retrain Tess or create new responses so that colleagues could see noticeable improvements in Tess every week.
A few months after launching, it became clear that some colleagues were using Tess similar to a search function and they expected it to be able to find any page on the intranet. We needed to find a solution that would help these colleagues, without having to manage all this content on both the intranet and in Tess. We decided to develop an integration with Tesco’s SharePoint, allowing Tess to search the site when it did not understand a query. We built a new fallback flow where Tess would ask the colleague to enter a keyword so it could search the intranet and return the top five page results. On the website search, page results are split across two tabs and often results from the second tab are missed by colleagues despite being equally relevant to the search term. We implemented logic within Tess’ search to always return a mixture of results from both tabs, ensuring colleagues see the top results for their search and therefore enhancing, rather than just mirroring, the existing search functionality.
Results
Tess is trained to answer over 150 queries and handles most conversations successfully by guiding colleagues to the relevant information. Tesco continues to use EBM to monitor Tess and we work together to identify key areas for improvement.
“The EBM team share the same passion as we do, keeping the customer at the heart of everything we do. Although the functional side of a chatbot and how it’s integrated into your business is vital for it to work, the team never lose focus on what really matters, which is the person interacting with the bot.
Liisa Brade Capability Manager | Tesco PLC
The EBM console is easy to use and the customer-influenced roadmap for enhancements keeps it moving forward. We are proud of the colleague-facing chatbot we have created with EBM and have plans to have the bot answering a wider range of queries in the future.”
