Last year I talked about how Browser Extensions are changing the way we work. The ability to reduce the number of places we need to go during our work day can directly improve company performance metrics and help business scale faster and more efficiently, all while creating a better experience for your customer. Some examples include:
– Minimizing the time it takes to reply to a customer support issue reduces your issue resolution rate and improves NPS.
– Answer prospect questions without leaving your inbox. The faster you can respond to inbound leads, the more likely you are to connect.
– Know when your prospects are engaging in the content you send them to better understand their interest.
Slack launched their platform in December 2015, allowing anyone to develop a bot on top of Slack. Since then, the number of bots for the workplace has exploded, creating a robust ecosystem of ways to work faster and more efficiently. Additional platform offerings were announced as well from Facebook, Microsoft, and most recently Google, creating the potential for a good ecosystem of new bot-driven experiences at work. Like browser extensions, bots live where you work, providing exciting new ways for teams to transform the way they work.
Bots are not without their challenges
Bots are steeped in the upward slope of the Hype Cycle these days, and many like to talk about the failure of bots. However, when you boil down most of the negativity, it has more to do with the scope of the problem, rather than the bot itself.
For example, we can’t (yet) expect a bot to “replace” a human. And, while I won’t conjecture whether or not that will ever be possible, the fact of the matter is that most bots don’t succeed because they go far too wide on the problems they are trying to solve. Because bots rely on NLP (natural language processing), like other fields of AI, you are only as good as your training data.
There is a direct corollary between the breadth of the problem you want to solve, and the amount of high quality training data you have access to. Many of us have probably experienced the limitations of some sort of broadly defined “personal assistants”. We try to treat the bot like an assistant, and then it doesn’t work, and we get mad! It’s not the bots fault, even though I want to blame “it”. It’s a mile wide and an inch deep problem, where the bot just doesn’t have a chance for success because it is being asked to do too many things. Much like a human and the way we learn, we don’t simultaneously try to teach ourselves to code, to sell, to do customer support, etc. We tend to optimize and get really good at a relative few things. Same idea with…