Navigating Parabol’s AI Roadmap

Open Questions

The key to great AI features lies in empowering users to build on their proprietary data created in our software. Parabol customers have great data from meeting history and task context. With the introduction of Parabol Pages, customers now have notes, documentation, and databases.

AI features have become ubiquitous. When we set out to discover what AI features would best serve our customers, we had a lot of questions:

What are the natural use cases within our existing software and roadmap?
What are customers’ ever-changing expectations around the use of AI?
What new software should we build from scratch as AI becomes widely adopted?

We reached for our process of discovery and rapid testing of assumptions through small iterations. We’ve reached several milestones:

Enriching meeting discussions with prompts and related topics
Automating meeting summarization and highlighting trends for team leaders
Enabling leaders to generate shared learnings across the organization

Enriching discussion topics

Team leaders have learned to expect more engaged team discussions in Parabol. This is part of our mission. How could we increase the engagement here, but also deliver new value?

We introduced discussion questions and related discussions.

Discussion questions help team leads kick off a topic. Related topics help teams stay aware of trends that might need their attention over time.

Under the hood, we developed an embedding service on meeting data which has proven to be a valuable long-term investment for LLM-based features developed later.

While these messages are automated, they are not interactive. We expect threads will be a natural place to interact with an agent in realtime further down the roadmap.

Opting in to automated summaries and insights

Team leads tell us how much they like our meeting summaries. These snapshots help when looking back and planning ahead. They share them with other leaders. Team members who are absent stay caught up. But, they require a bit of parsing for the context. You can see discussion topics, reflections, tasks, and go back and look at threads. It was clear meeting summarization would be a win here.

What was less clear is where and how to expose trends across meetings. We made this the goal of an in-person design sprint in San Diego. We developed some prototype tooling for generating insights from meeting data. And we had a little contest to see who could deliver the best insights from some smart prompt engineering. We had some judges from external friends in tech and consulting industry gauge our team insights.

Another challenge was gauging customers’ comfort level with AI features. Sometimes meeting data has sensitive topics. Who would have access to these insights? We tested carefully with team leads, reaching out to opt into this beta program, and provide feedback.

We started by providing static team insights reports with a prompt interface soon to follow.

Embedding cross-team insights in Pages

With the confidence gained from feedback around team insights, we were keen to introduce AI Insights in the primary user journey.

We built Pages so teams could both provide context to and capture knowledge from meetings. We embedded the AI insights and summarization directly in the experience.

The previous meeting summary artifact was static, including the LLM generated summary.

Now when teams conclude a meeting we generate an interactive summary with insights leading the page. They can change the prompt. This is a firsthand way to discover and learn how to use the insights block and the page blocks. The pages include topics, tasks, and an export of the meeting.

In the future, we want to include page content, databases, and transcripts to the insights context.

Next Steps

We are discovering our next AI features from 0 to 1. We are shaping the following features on our roadmap. We are…

Creating simple ways to connect to Parabol using OAuth API and MCP, and personal access tokens. This will allow teams to explore the data in custom workflows without any UX constraints.
Exploring interactive chat in Parabol for immediate access to the data without setting up a custom workflow.
Expanding the context to include pages, databases, and transcripts.

Internally we are focused on integrating agentic workflows in a way that allows us to make the best decisions and do our best work. Building for ourselves first is key to our future vision.

Bonus Round

Our Icebreaker round is a popular feature. Teams have the option to generate unique flavors of questions from our prompt library.

AI is great for automatically grouping and theming topics.