Publishing company deploys Alexa chatbot in record time
Our client is a medium size UK business. In early 2018, they decided to invest heavily into voice. They strongly believe voice is the future of search and, thus, the major contributor to their future bottom line.
One of their executives contacted us for help. It became clear that, to properly champion the technology, he was going to need something tangible to show the board.
Settling on an MVP
Championing chatbots is interesting, especially voice chatbots. You can get extremely excited talking about it and be met with a blank stare from your audience. It is only when you pull out an actual chatbot and get people to try it that their eyes light up — they get it.
For this MVP, we needed to pop open the eyes of a board of executives. What would make them tick? We couldn’t build an entire complex bot. It would blow their mind, sure, but it would also take months. We had to build something in a few days that could raise enough adrenaline to get the go-ahead to carry on.
We settled on a simple demo: provide a line manager voice access to their employee’s last six performance reviews.
Turning employee records into conversations
With only three days to put this together, we had to fudge it a little bit. We had no time to connect any APIs or create proper integrations with their HR system — that's the easy stuff and wasn’t the point.
First, we requested an anonymised extract of 200 unique performance reviews. We used this data set and a simple natural language understanding model to spin up over 15,000 performance reviews for 1,500 fake employees.
With this, we were ready to develop some conversations. Using Alexa, we started building one conversation flow to allow a line manager to request an employee’s last performance review. Good, but not enough. Next step, we allow the line manager to ask for any of the last 6 performance reviews. Then, we allowed them to ask for any negative performance review (any geeks would know it's pretty tricky to do this at scale in just a couple of days!).
Finally, we allowed the line manager to request a performance review, listen to it in full (through Alexa), and organise a meeting with that person’s calendar to organise the next review.
Now we were happy.
Something tangible and a roadmap
Sat quietly in the corner of the room (in case the proverbial hit the fan!), we had the chance to experience the test first hand. After weeks of trying to champion voice within the business, our client now had something to try. Testing the demo proved fruitful. It truly displayed the power of voice and its ability to easily access data in a conversational way.
This was all they needed. ubisend is now working through the first, real iteration of this build. We’re finally replacing the dummy data, fake names, fake reviews, and fake APIs with the real deals. Using the case study from the MVP, our client has managed to start advocating customer-facing voice chatbots.
Due to the confidential data we help our clients manage, we often work under hefty NDAs. If you're interested to know more about this project, get in contact, as we may be able to introduce you to the client for a chat.