Behind the Scenes: Mattermost OpenOps AI Mindmeld | August 15, 2023

Mattermost

Aug 15, 2023

Tune in for a behind-the-scenes discussion on the advancement of Mattermost's AI tools and how they're being integrated into the team's current projects.

In this installment, the team explores the bot's capabilities — including how much it can remember — and evaluated whether they should focus their efforts on improving the AI model or adding more features to the AI plugin, among other things.

To learn more about how to get involved with the Mattermost OpenOps AI community, check this out: https://forum.mattermost.com/t/welcome-to-the-mattermost-openops-ai-community/16144/2

Summary:

The meeting discussed several key topics. First, a community moderator named John Combs has created a chat GPT support bot using Zapier which acts as a technical support rep. The bot has been well-received, and there was discussion on whether to make this information public. A potential demo of the bot is being scheduled and it is also suggested that Combs be invited to a future meeting.

There is also an ongoing exploration of the bot’s capabilities, especially how much it can remember. There was a conversation about the potential costs involved and if these should be covered if the tool becomes widely used, particularly since Combs might be using his personal account for the project.

The team evaluated whether the focus should be more on improving the AI model or adding more features to the AI plugin. They settled on the latter, pointing out that they’d mostly covered the good models and wants to move towards customization.

A demo by Jesus about a new functionality extension to the ‘unread message bar’ was shown. This feature allows users to summarize unread messages in a channel. While the frontend has been developed, the backend is not yet implemented, but a pull request has been made for its functioning.

Finally, the team discussed a potential UX issue regarding the inconsistent appearance of the ‘unread message bar’ and needing another way to trigger it. The team agreed that this could be tackled through plugin extension. The meeting ended on a positive note, with participants appreciating its insights.

Key Discussion Points

GitHub Support Bot Discussion

Community moderator John Combs created a V1 chat GPT support bot using Zapier.
The bot, Matterbot, simulates a technical support rep and tested well in initial use cases.
The bot is not perfect and requires improvements.
The bot’s response accuracy and possibly retaining state were impressive, providing relevant responses in the demo.
There are concerns about making the bot public and who will bear the cost if it significantly increases usage.
It is believed that Matterbot could be useful internally, serving as a draft for the support team, effectively reducing their workload.

Action Items

Invite John Combs to an upcoming meeting.
Explore how Matterbot can be improved.

Identifying Next Priorities

There is a suggestion to try out different models to heighten the excitement around the AI plugin.
Some team members believe that adding new features and customization paths would add more value than exploring new models.
There is a need to gain access to the AWS Bedrock model.
A decision must be finalized on what the next highest priority should be.

Action Items

Continue discussion on the next priority.
Make attempts to gain access to AWS Bedrock.
Invite John Combs, the creator of Matterbot, to participate in the discussions.

Unread Message Bar Updates

A demo of the newly added ‘Summarize’ functionality to the unread message bar was shared.
The ‘Summarize’ button presents a summary of unread messages.
The back-end functionality has been implemented successfully and needs to be connected to the front-end.
There are some User Experience (UX) inconsistencies with the presence of the unread message bar that need to be rectified.

Action Items