How to make an award winning Housing Sector chat bot using Microsoft Power Virtual Agent

Crafting an Award-Winning Housing Sector Chatbot: Unlock the Top 50 Must-Have Topics

James Berger
4 min readNov 5, 2022
From Left to Right: David Matthews, James Berger, and Michael McDonagh. 🏆 Celebrating our achievement: Housing Digital Innovation Award for Best Online Service 2022! 🎉🌐

No matter what industry you’re in, whether your company is big or small, there are few things more important than customer service. It’s one of the key factors that can make or break a business. And with customers expecting more and more 📈 from businesses every day, it can be a challenge to keep up.

That’s why conversational AI 🤖 has become such an important part of modern customer service strategies. They allow companies to enhance the customer experience while improving efficiency levels by minimizing repetitive and redundant tasks.

Following on from my previous blog post, where I first introduced how Origin Housing had begun using Microsoft Omnichannel and Power Virtual Agents. This post is a year on and delves into what I have learned 🧠 from the experience and how you can shortcut your way to success, providing an experience customers will love.

Understanding customers

It takes time ⏳ and a lot of data to really be able to understand what it is your customers are asking for. Even when you initially think you have captured all the different ways someone might ask for something, there is always a phrase for which you didn’t account.

As an example, we have 685 different ways where someone has said they are interested in a property. 🏠 There are the expected phrases such as “looking for a house”, “need a place to rent” and then you have to account for all the less expected phrases which include unintended spaces in the middle of a word and typos. These all need adding as a trigger phrase, in order for the chat bot to recognise and understand next time.

This graph shows month on month the volume of conversations and the % of conversations that the chat bot did not understand and had to escalate.

It took over 14 months to reach a stage where Basil Bot understood more than 80% of the conversations in a given month. This has resulted in customers not having to reword what they were asking for and less conversations being escalated to a Customer Services agent.

Topics

The topics most requested by customers are log a repair, account balance look up, speak to a neighbourhood manager and interested in a property.

📓 Here are the top 50 topics that Basil has been built to respond to.

It is probably not a surprise to see that the most frequent conversations that Basil has are about logging a repair. Following the ‘Report a repair’ topic, customers speaking to Basil are requesting their ‘Account balance’ and arranging to speak to their ‘Neighbourhood Manager’ (Housing Officer).

The phrases associated with the topics, took time for us to develop. We had to have a lot of conversation 💬 transcripts in order to work out what it is the customers firstly wanted and then secondly, create a great chat experience. When I say a great chat experience, we wanted Basil to be deliberately conversational and not to just reply with a link to a page on our website (FAQ style).

Supporting Customer Service Teams

Chat Bots are 24/7 ⏲ and they can help reduce the morning backlogs that occur from enquiries piling up. There is however a misconception that chat bots are busiest outside of the core working hours. Basil Bots peak periods are Monday to Friday, between 8 am and 6 pm. To some this might not be a surprise, most of us still assume that Customer Services are only available between traditional working hours.

Basil Bot supports both the Customer Services team and the customers. Basil can resolve up to 78% of conversations. This provides the bandwidth (headspace) 🧠 staff need in order to deal with more complex enquiries that need human intervention.

85.46% of conversations are during the working day and the majority of these conversations take place between 8 and 6.

Basil Bots key results

🤖 1,549 repairs logged
👩‍💻 22,370 chats to date
👩 976 Neighbourhood Manager appointments arranged
🔥 78% resolution rate
💬 1,798 conversations via WhatsApp

So that’s what it took for Basil Bot to win ‘Best Online Service’ at the Housing Digital Innovation Awards 2022! 🏆 It was a great achievement personally, being recognised for all the hard work that had been put in developing Basil Bot. And even more satisfying to know that customers are truly getting a Chat Bot service that works.

Reality

Customers in the long run, will still demand and need human interaction and one of the most critical aspects of any Customer Service team is the ability to have meaningful, empathetic conversations with customers. It’s not enough for a chat bot to just answer questions and provide some self-service actions. Customer Services staff will still be needed to be able to empathize with people who are frustrated, upset, and angry.

The reality for Basil Bot is that the ability to understand and deal with all the ways a customer might ask for something is still being refined and will take years. Even further away is not being tripped up by tricky questions and complex language.

Want more like this?

Let me know in the comments below if you want to see more content on Power Virtual Agents and I’d love to hear how your chat bot projects are going.

--

--

James Berger

Enterprise Architect | TOGAF Practitioner | 11x Microsoft Certified | Dynamics 365 & Power Platform | Scrum Master