The CEO of a prominent financial institution looked aghast as I described features of various popular chatbots — “I’ve gone to such lengths to protect my time. Why would I let a piece of software interrupt me?” was his very valid complaint.
Paul Graham’s classic post on Maker v/s Manager schedules is as relevant today, as when first written — perhaps even more so, given the multitude of interruptions we all battle. Many first generation chatbots incorporate auto-notifications and pre-set reminders. Some are extremely useful, but also annoying. In some cases, the annoyance is a feature, not a flaw — as I explain below. But, as we attempt to create “digital co-workers”, shouldn’t we aim for the strong, silent types?
First, lets define the two types of chatbot initiated messaging prevalent today. I will use the example of Howdy (a prominent Slack bot that we love), to distinguish the two:
– Reminders occur when the chatbot preemptively interrupts what you are doing (typically at a set time) to check in with you. An example is when Howdy pings you and asks for status on various projects. The schedule is typically set by the Slack admin for the entire team. The purpose is to prompt a response from each team-member. Inevitably, I find myself in a situation where I am unable to respond at the moment. Granted, there is a window for response, but I am drawn to the ping, don’t respond — and then drawn back, when Howdy warns me that my response window is ending.
– Notifications are pieces of information passed to you, such as when Howdy compiles a digest of all team-member statuses. They can be viewed immediately, OR at a later time. But the ping, inevitably draws me to view the info right away.
This partially attenuated attention is the exact opposite of what is required for Maker mode — uninterrupted time for deep work.
Is there a better way? The lens through which I searched for solutions is from building Kelasa, a Slack Bot that extends Workfront (enterprise project management software) to Slack.
Power to the people: Chatbots can and should be individualized. Each person has a different working style and settings at the admin level cannot satisfy each individual.
Suggested responses and menu driven interactions helps users set up their individual preferences. A wonderful example is Kip’s Slack bot, that provides Settings access to all users, in a way that’s cleverly hidden within extended menu buttons.
Alternately, Setup and user preferences can be part of installation or as a first-run feature. This approach also works when there is a response window i.e. if you need to collate all individual responses before a particular time. Conversationally, the user can be driven to pick a time preference within the window. On platforms that allow web-views, directing end-users to the settings website is also a possibility.
Sometimes reminders are deliberate. For example: Kelasa includes the ability to set up a reminder to ask users update status and time on their project tasks — similar to that nagging email from project managers (except that the bot takes over the tedious task from the PMs). SlackBot reminder messages also fall in this category. Obviously, these need to be interruptions, since they are designed to evoke a user response.
Notifications that arrive at random intervals need more careful thought. Some individuals might prefer all of their notifications to feed into a single channel (or as a digest), for a quick review later. For platforms that don’t include channels, allow for preferential treatment of certain types of notifications.
Allowing individuals to determine when and how they are messaged, is a powerful way to allow them to set boundaries.
Incorporate machine learning: If I snooze a reminder each and every time, shouldn’t the chatbot “learn” my behavior and adjust accordingly. Incorporating simple heuristics, based on user action and context, can help train chatbots responses and customize them for individuals.
The beauty of the conversational UI is that its remarkably unpolluted. Paying attention to the way notifications and reminders work, will help keep it that way.