Second-order features

In logic and mathematics first-order logic quantifies only variables that range over individuals; second-order logic, in addition, also quantifies over relations.

Thankfully for the both of us that's as deep as we are going to get on mathematics in this post. Instead I'd like to extend this concept from mathematics to software feature development.

Let's start with what I consider first-order features. For us here at Connie AI that would be the AI assistant chat interface. We aim to build the most accurate and reliable question answering interface directly into your Confluence spaces. We've written more about the details of that here: https://blog.alu.ai/rag/

The thing with second-order features is that they are impossible to build until you have robust first-order features. Oscar and I realized that Connie AI as a chat assistant interface, while useful, was what we call a single-player feature. As a Confluence user, you come to Connie with a question, get your answer or not, and continue on with your day. What was missing was a way for Connie to help not just the individuals with questions, but the team as a whole. In brainstorming we came to the realization that because of what we had already built Connie had everything it needed to be even more of a team player.

In order to answer questions about a space well Connie AI builds up a robust index regarding the content of a space. We realized that what we also had was the questions that people are asking of a space, as well as all the metadata around those questions. Things like which documents were referenced when answering, which people we considered likely experts, and crucially whether Connie was able to formulate an answer to the question. With this information in hand Connie can play a leading role helping teams understand how well their spaces are doing. Connie AI's new Insights feature is our first second-order feature. We take the information Connie knows about the space and the questions being asked and display that to the team.

Our hope is that by providing these insights to you and your teammates it will provide a useful starting point to identify gaps in the content of the space that could be improved, making the space more useful.

This is very much just the starting point for Connie AI's space insights feature. We already have lots of ideas on how to make it smarter, more useful, and more targeted. For unanswered questions, Connie could provide pointers to which documents might make sense to update to answer the question for example. Connie could also perhaps use its real-time knowledge of the Confluence space to automatically mark unanswered questions as addressed when pages are updated.

We'd love to hear your ideas on how this feature might be useful to you and your teams. You can reach out to Oscar and I at contact@alu.ai

If you'd like to be notified of new posts and announcements from us about Connie AI and future products you can subscribe below: