Pan Macmillan needed help to create more opportunities to engage with their customers – offering ideas on which books to buy as a gift, or to read themselves.
We developed a personalised chatbot recommendation assistant, that would help choose that perfect book – whilst reducing one of the most common customer support queries received on their social channels. (ie; ‘what the hell book do I buy for my significant other/friend/colleague/pet’)
The requirement was to have an automated (but personal) assistant that could work 24/7 to answer questions, communicate with users and offer smart recommendations on their next book gift or read.
The campaign was directed at people looking for just the right book gift. It featured multiple API integrations, NLP, and e-commerce integration to make it really easy to find that perfect book.
So we developed a book recommendation engine, available on Facebook Messenger, featuring automated ideas for great books to read.
The addition of expert-curated content also helped make the experience a bit more human and made sure the bot could answer book-related questions clearly. It could also respond in a personal way if it didn’t understand what it was being asked, this is called smalltalk in the conversational AI area – and if done well, can give bots more of a personality.
Personalization
The key to a successful chat campaign is making sure the use cases are watertight, and the bot fulfils its purpose elegantly and with relevance.
The Pan Macmillan chatbot was built to automate the recommendation process for the user (while under the hood there was a complex set of calls from different APIs to combine).
Understanding that each reader is different is important, and tailoring the experience for each individual makes better and more relevant recommendations.
Integration
We pulled data into the chatbot from two of the main APIs – Supadu and Goodreads, alongside data from Pan Macmillan’s own data sets. Amazon integration was also incorporated, making it a snap to go and purchase the recommended reads.
With a lot of books being published, the chatbot needed to be smart AND personal – simplifying the vast amount of possible choices to ping back relevant, well-ordered recommendations.
We supported this with guided categorisation by genre, author, and release date to make the experience relevant and simple.
We used the Supadu API to pull in book meta data to the chatbot
Synthetic are a smart bunch of emerging technology experts, and nice team to work with. We've worked on some innovative projects in conversational AI and voice, and would recommend for any digital projects in that space.
