Top must-have features of AI-powered Slack Bots

As Slack slowly, but surely, challenges (usurps?) email and common word processor productivity suites as the operating system of work, it loses its focus on the core messaging functionality and expands to many more roles. Some of those additional roles are:

  • External messaging (as a secure communication channel for partners and customers)
  • Knowledge sharing
  • Support issue management (IT and internal team support)

Most importantly, Slack is the medium though which the collective knowledge flows through the organization. Whether that knowledge is related to finance, product, engineering or just the free-flow of ideas, Slack enables the exchange of all of it Within that flow of conversation there is a massive opportunity to leverage this data. Enter the Slack Bot; the perfect vehicle for analyzing conversational data and providing actionable instructions to improve the effectiveness of self-serve support channels. The top must have features in any Slack Bot that enable better self-serve support are:

  • 🧠 Sensing frequently asked questions in Slack
  • 🏥 Triaging support in Slack
  • 🔍 Collecting data on knowledge gaps in Slack

Sensing questions in conversations in Slack

When open questions are asked in a Slack channel, an internal support issue is being opened and the support seeker is requesting assistance to unblock their work. An AI-powered Slack Bot can be tasked with identifying question patterns in Slack and providing some form of assistance. When a Slack Bot is equipped with Natural Language Processing (NLP), it can be trained to identify a question pattern and provide some level of support. Some examples of beneficial suggestions might be:

  • Relevant wiki resources related to the support request
  • Relevant files, documents that may be helpful
  • Giving the opportunity to crowdsource a wiki article to resolve a knowledge gap (if one exists)
  • Offering other self-serve support workflows such as ticketing or creating an issue/bug in a tracking system

We typically don’t want an NLP-powered Slack Bot to interject all the time, though. For example, when someone asks a common question in Slack like “Does anyone want to go for lunch?”, we don’t necessarily need our Slack Bot to provide support resources (unless maybe the Bot finds some popular local take-out menus 🌯 🌮 🙌 😀) – it may just seem like noise at that point. So the training the Slack Bot is going to be important to ensuring that it when it does have something to say, its at least intelligent!

Triaging support issues in Slack

When a support issue arises, there is an optimal workflow that an AI-powered Slack Bot can execute to guide support seekers through. Individual needs may vary, but the optimal triage might look something like this:

  • 🤖 Sense a question asked in Slack (with AI-powered NLP)
  • 🔍 Conduct a real-time search the wiki or other repositories for relevant resources (in the background)
  • 📈 If confidence is high (>80%), privately provide one or more links to resources – the private suggestion is required by Slack as a design requirement so as not to create excessive channel noise
  • 🤝 If the provided resources are helpful and resolves the issue, we have success and we can give the support seeker the opportunity to share the knowledge with the channel publicly – in the background, the Slack Bot should record this event as a deflected ticket 🥳
  • 🎫 If the provided self-serve support resources are insufficient, the support seeker can be provided the opportunity to independently open a support ticket so that the issue can be escalated for resolution – in the background the Slack Bot can log the issue as a knowledge gap and create a request for documentation to be created for the next time the same question is asked
obie suggest create jsd ticket in slack

Collecting trends and data on knowledge gaps in Slack

One of the most painstaking recurring processes that occur for those that manage knowledge is the collection of data relating to the discovery of knowledge gaps. With almost all recorded conversational data being stored in Slack, it is the perfect data source to capture repeated internal support questions. Without a tool like an NLP powered Slack Bot, discovering knowledge gaps, say monthly, would be a time-consuming, low-value exercise and would only get amplified the larger the company gets. Ultimately, the true value of a Slack Bot is in its ability to recognize, log and present this data into actionable tasks. Then the knowledge manager can assign authoring tasks to subject matter experts to ensure that the most accurate data is available org-wide.

Obie Dashboard and Analytics

Slack Bots can be a powerful addition to the organizations that are heavy Slack users. Like any AI-powered technology, the main constraint is the volume of data available to the bot and the training time that is invested to improve effectiveness and accuracy. If your organization is fairly large and depends on Slack for internal support, an AI-powered Slack Bot might be