ActuateBots – Simple idea, powerful outcomes
Think of a Thursday morning. It is very likely that even before you reach your work you have already interacted with one or more of Alexa, Google Assistant or Siri – for the weather, for directions, for the routine call to your parents, maybe more. Then after lunch perhaps you permit indulgences like browsing through latest fashion trends, or maybe you punish yourself and finally get done with the chore of rebalancing your portfolio. Each of the websites has chatbots that take you some distance. Then the evening at home – songs over Alexa, wine ordered via Bevmo chatbot, next day’s weather forecast from Siri – bright and sunny!
Ever wonder why you do not have these experiences at your work? Many organizations do have chatbots, even if text-driven, for things like customer support, HR, vacations, etc. (If yours doesn’t, do something – these chatbots are good). Still, these chatbots only educate or confuse, and nothing more. A typical executive is left to their means when consuming knowledge or doing mundane tasks in the organization.
In 1999 management guru Clay Christensen published his seminal “Jobs to be done” framework. His research shows that users consume a product or service because each of them has a job to be done for the user. The job may not be what the product or service is ostensibly about. For example, Prof. Christensen’s team found that a dairy used to sell a large number of milkshakes between 7:30 and 8:30 am on weekdays, a not very intuitive time for milkshakes. Turns out, the job the shakes were doing for commuters was to blend breakfast with something to do on the way to work.
People come to your chatbot because they need a job done. In contrast, traditional chatbot approaches tell the user how to get the job done, at best, and then stop. If the bot got that right, the user now must figure out the next steps on their own. If not, then they have to restart the chat, think of asking the same question differently. Let’s face it; no chatbot is 100% accurate.
Let’s take an example. An oil and gas major has developed a robust Standard Operating Procedure (SOP) for maintenance of equipment. Given the tight placement and interdependence of equipment in oil wells and rigs, the SOP requires many approvals and checkboxes. It is easy to believe that some frequently used maintenance schedules were massively delayed waiting for necessary approvals.
The team built a bot over the SOP, hoping that once engineers can navigate through the SOP in an interactive way, things will improve. They were right, but that improvement was dwarfed by the simplicity of processes once we made the bot automatically fire request emails, reminders and API calls for necessary parameters. The bot does not only preach, but it also professes.
In most bots, the active interaction is centered on real estate worth a couple of lines. There is a lot of white space just not being used on the entire screen – plenty of it, even on a mobile screen. In fact, designers have to think of creative ways to not leave that portion simply white (or black).
So here is a simple idea. You already know the intent of the user; you figured it out while trying to answer their question on the bot. You already have the white space as well. Why not put in a button or a form, that gets user’s job completed? Why not put three, because the bot is much more likely to get the intent right in three chances than in the first. There you have it – an ActuateBot.
An ActuateBot is a chatbot that thinks like a human and gets things done.
Let’s make this real. Scientists at a pharmaceutical company interact with a large corpus of highly technical documents using a chatbot. They ask things like “what is the chemical structure of Toluene,” and the bot returns some diagram. While designing this bot, we were pretty sure that the scientists are not asking this question just to be reminded of how beautiful in its simplicity Toluene is. There are multiple jobs to be done – send this to another user, add it to a report with all the necessary metadata, investigate which synthetic routes are most appropriate, query for variants, and many more. We sat down with our clients and ended up creating an ActuateBot that saves them a lot of time while drafting reports for the FDA.
The user experience with ActuateBot is very different than regular chatbots. Getting tedious tasks completed by a single click or tap leads to real customer delight. The perception of accuracy is much higher because the user is not only looking at the first answer by the bot, but also the actuation levers on the screen. The experience is far more enriching.
In our experience with ActuateBots, we discovered yet another insight. Users trust a chatbot or a human with sensitive information far less than an internet version. For a financial brokerage, we implemented an ActuateBot to go through the top news of a day relevant to every portfolio. One of the actuation is to enable trade. We found that people were more comfortable using these templates than typing “sell all IBM” into a chatbot.
From a data science point of view, this is super valuable as well. ActuateBots create significant opportunities for your AI to train. Every time a user clicks on the second or third actuation, there is a strong signal to be fed to machine learning. Insiders know that these signals are rare otherwise. We used these signals to train a contract management ActuateBot for a construction company to very high accuracy. It would have been difficult otherwise.
Simple ideas lead to powerful outcomes. Surely, soon enough, workplaces will have people talking to each other or a computer. A few taps here or there. A lot of thinking. A lot of strolling around. A lot of smiles – all powered by ActuateBots.
Simple ideas are also hard to implement sometimes. Let us know if we can help in your ActuateBot journey.
