Revolutionising Customer Service: how Bouclair leveraged AI with UPEO’s CLAIRE

Bouclair, a leading Canadian lifestyle brand, needed a way to enhance its customer service efficiency without compromising data security. Working with UPEO, they found a solution by integrating Artificial Intelligence (AI) into their customer service operations. This collaboration led to the development of CLAIRE – an AI framework designed to assist service desk agents with quick responses and efficient customer interactions. By using AI to streamline their processes, Bouclair was able to automate tasks, improving the efficiency of their customer service.

Bouclair is a Canadian lifestyle brand with 60 stores from coast to coast and a popular online shop. Founded in 1970 in Pointe-Claire (Quebec), Bouclair designs and sells furniture, bedding, home accents, wall decoration and seasonal products, and has over 1,000 employees. With its slogan of ‘Home starts with a feeling,’ the brand attaches great importance to customer experience and customer service. Enhancing customer service is also the mission of the digital operations team, that automates as many processes as possible to give the shopper an integrated end-to-end shopping experience, whether they buy online or in one of the stores. “Over the last few years, we have digitised and integrated most of our processes to enable us to work omnichannel,” says Jean-Sébastien Ré, manager digital operations at Bouclair. “We can now work more streamlined, and having all the data in a system of record makes it easier to interpret that data.  All departments are reporting gains in efficiency. Order fulfillment, for instance, is now a matter of hours rather than days.”

To reach its goals, Digital Operations is always looking at integrating new technology and, just recently, Bouclair started using AI to assist the service desk agents. “Many of our service agents are part time workers and we regularly see new hires. But we have a lot of different procedures that are complex, which an agent needs to learn in a short period of time. When we have a peak, for instance at Black Friday, we need to hire new people, train them and tell them in what different places they can find the relevant information. That’s not easy and we were looking at a way to support them using AI,” says Jennifer McKernan, systems analyst at Bouclair.

Connecting Salesforce with OpenAI

Previous attempts to let AI solve the challenge in Customer Services were blocked by the reluctance to open up the database to external systems like AI, as there is personal information involved. This was solved by the co-development of the AI framework CLAIRE by both Bouclair and UPEO. CLAIRE sits between Bouclair’s Salesforce data and AI frameworks /large language models. CLAIRE uses the AI engine of these models, but shields the customer data from them. UPEO wrote a custom API to connect Bouclair’s Salesforce clouds with OpenAI.

The solution for Bouclair was co-created by the digital operations team and UPEO, with whom the COO had worked together previously. “I see it as a joint venture: UPEO developed the puppy, and we trained it,” is how Jean-Sébastien Ré describes the co-creation process. “As AI and LLMs are still very new technologies, and the models and frameworks are being updated and improved constantly, it’s really a process of trial and error. I think we experience the same as the people who built the first websites, first using HTML, then discovering PHP and CSS and having to adapt the process, all the time.”

CLAIRE has the answer

The project was kicked off in January 2024 with just three key users working with CLAIRE but, after a couple of months the user base was expanded to the entire service desk. When a customer submits a case through the web form on the website asking, for instance, ‘where is my order?’, CLAIRE will retrieve the order, consider the different objects related to that order, look at the internal procedure related to it and come up with a suggestion of what the agent can do for the customer. CLAIRE also writes a draft email that the agent can send to the customer. The agent reviews the solution and the message, adapts it where necessary, and sends it to the customer. “After just a couple of months of using CLAIRE, we are currently at between 40 and 60% correct answers, and we are aiming to get it to 80% correct answers,” said Jean-Sébastien Ré.

Educating CLAIRE

Large language models need to be fed with data before they can function effectively, and Bouclair found out the hard way how important the quality of the input is. “We fed the system 200 pages of documents with our procedures,” Jennifer McKernan recalled. “But the more data we put into the system, the dumber it got, causing it to ‘hallucinate’. And that’s because there were inconsistencies in our basic data.” Reviewing the material that Bouclair provided to train CLAIRE, they found information that contradicted other pieces of information. This confused CLAIRE. “That is one of the important lessons that we learned: you can’t just throw all the data at AI and assume it will work perfectly. You have to provide it in a more structured way.”

To get CLAIRE to find all the necessary information to do the job properly, the Bouclair team spent a lot of time on mapping CLAIRE to the custom fields in the CRM system.

Another lesson learned is providing the system with feedback faster, and set up KPIs. Bouclair now asks the agents to rate the suggested solution with a ‘thumbs-up/thumbs-down’ and inputting more information. “With the benefit of hindsight, we should have started this process earlier,” Jean-Sébastien Ré admitted. As working with CLAIRE is still pretty new to the services team, Bouclair needs to keep reminding the agents to provide the feedback.

Looking forward

Now that Bouclair is getting more experienced in using AI, new potential use cases spring to mind. CLAIRE is currently only used by Order Management and CRM, but Jean-Sébastian Ré can already see it being used by the finance department, who also get questions from customers, and even in the Bouclair stores throughout Canada.

The Bouclair team are very pleased with the way UPEO worked with them. “The solution they offered to us was pretty stable and they were always very responsive,” said Jean-Sébastien Ré. “Sometimes we broke things and needed UPEO to fix it, or do some extra coding. I found they listen well to our feedback, concerns and questions. They answered us promptly, despite the time difference, and actually provided us with good ideas on how to better analyse the data in our KPIs.”

Press release featured on Belgium Cloud and DVO.