Last week we hosted an event for our network and friends: bots x beers. Really, the title explains it all: we wanted to have some beers and talk about bots. We invited some very knowledgeable people around to come and share their experiences with bots with our audience over some pizza and beers.
Read on to find out more about what their conclusions were and what that means for bots in the enterprise.
Why are we talking about bots?
At OrangeTrail we are excited about what bots could mean for the enterprise: how can we use these tools to make the lives of employees better? That’s what we like to think about - and we know that bots can make the difference: they have higher read & response rates than any other enterprise medium, and have the potential to improve processes in a big way.
What did we learn?
Insight #1: Start with one thing
“Do one thing and do it well”. James pressed on this point as he gave his presentation: rather than trying to do everything and solve all the problems you see, try to start with one problem and finding a solution for that. Likewise, Phil mentioned “start small”, again mentioning that it’s too easy to try to do too much at once.
As tempting as it can be to go down a rabbithole of ideas and trying to include everything in your bot, all the panelists made it clear: start small, choose a problem to solve and focus on that, and on that only.
Insight #2: Smart bots need time and labour to learn
As our speaker Fadoa put it: there are scripted bots, bots that walk you through a pre-written flow of messages and smart bots, bots that are able to understand you and learn. However, making a bot understand what you’re saying takes more than just flicking on the NLP (natural language processing) switch. Having the bot being able to parse and “understand” human languages is only step one. Once there, the bot needs to learn the context in which the conversation is being had.
Take the word navy: it can be a color, but also a military entity. In a normal conversation you’d know which one I’m talking about but how would the bot know? It needs to be taught these contextual queues. James also shared a bit about the process at booking.com: there they had CS employees tagging messages with topics in the first weeks after the bot was live: this way it learned to link phrases to meaning.
Dennis’ experience also reflects this: he said machine learning takes time, and to not underestimate the importance of dashboards and the human factor. He also mentioned that one needs to choose the medium wisely: after all machine learning can also be based on voice, on images, you name it!
Insight #3: cool stuff doesn’t necessarily mean value
“Cool stuff doesn’t necessarily equal value”, Phil shared. It’s easy to chase the cool-sounding bots that sound like they can do a lot, but the reality is that the technology still knows it’s limits. Rather than being seduced by the opportunities that the technology brings, make sure you keep a focus on what the business needs. Have a look at the 7 areas where bots can make the difference that OrangeTrail identified.
Insight #4: the details matter
Ok, so you’ve got your business need, you picked a small problem to begin solving with a bot and you’re ready to commit to the time and effort needed for success: what’s missing? It’s time to really think of your bot as an interface.
Phil mentioned the parallel between the original command prompt interface and the move towards chatbots, with the main difference being: we used to have to speak the language of the computer while now the computer is learning to speak our language. With that in mind, what can we do to make the bot a better interface?
Think not only of buttons and images, but also of things like personality, tone of voice and choice of words. James and Phil both talked about this: while it’s important that the bot have some personality, it’s important that it doesn’t veer too far in one direction or the other. So think about who your audience is, what your bot is helping them with and how that should sound. Is your bot serious? Does he or she tell jokes? Is it gendered? What works best for your audience and context can vary!
Speaker #1: James Butler, UX Designer at booking.com
James is a User Experience Designer at Booking.com who has spent the last two years working on messaging and chatbots. He helped to design, test and launch the Booking Assistant - Booking.com’s chatbot assistant. Read about the process here.
Speaker #2: Fadoa Schurer, Founder and CCO at embot.ai
Fadoa has a lot of experience in building Dutch chatbots for customer care & e-commerce. Embot helps companies automate their conversations by creating virtual assistants and chatbots with intelligent skillsets, with a focus on the use of the Dutch language.
Speaker #3: Dennis Jungerius, productmanager at chatdirect
ChatDirect develops innovative communication solutions to work more efficiently. They specialise in the healthcare industry. When Dennis told me about the things healthcare bots can do
Speaker #4: Phil Kropp, partner at OrangeTrail
Phil is one of the owners and founders of OrangeTrail, and works closely with several of our clients to find ways to make their collaboration processes more effective.
So that was it: what I took away from the OrangeTrail BotsxBeers events. It was a fun evening, and I’m already looking forward to our next edition in April!