Inspiration
The inspiration for this project came from my own daily work. I am working as a Devops engineer for a bank. Daily, we have multiple calls and messages discussing about live incidents, testing environment issues, backlog items, emails to be sent, Jiras to be raised for us and for other teams, and the Confluence pages documenting the steps of work done/discussions done.
This increases a lot of manual activity. We need time to create Jira/Confluence pages manually which increases work for engineering teams.
OpsPilot is an AI workflow automation agent that monitors workplace messages and notes, identifies actionable incidents or tasks, and with human approval, it creates Jira tickets, sends Slack/Outlook alerts, publishes Confluence documentation, and prepares employee EOD summaries.
What it does
OpsPilot monitors workplace channels such as Slack and uses AI to identify incidents, tasks, and requests. It then prepares actions such as creating Jira tickets, sending triage alerts and emails, or publishing Confluence documentation. These actions are presented to user for review and approval before execution. OpsPilot also maintains an audit trail and generates end-of-day work summaries.
How we built it
For this project, I started with designing the product, architecture, workflows, and integrations. Once I shared this with Codex (5.6 sol Model), it had first build the entire codebase consisting of frontend and backend code. We then added Microsoft OAuth authentication through the Entra platform, configured trusted HTTPS certificates for local development, and connected Jira, Confluence, Slack, Outlook, and OpenAI using environment-based credentials. After configuring those values, it did the local testing and confirmed that the API calls are working fine.
It then helped me to host the website on Render and provided me the environment variables to be changed for hosting the website on Render. To make the application safer and easier to operate, we added human approval before external actions, employee privacy and consent screens, audit-friendly activity tracking, scheduled intake, and end-of-day summaries. Codex also helped in creating the automated e2e test packs which allowed workflow to be tested.
The project also includes a synthetic demo-incident feature that allows users to experience the workflow without connecting their own Slack workspace.
Challenges we ran into
Some of the challenges which I faced was when I was setting up Render environment for hosting the website. I had never worked on Render before and had seen few failed builds. So, I had to share the failed logs with codex to identify the root cause and update the code/environment variables to get the build and deployment successful. I had to also update the app registration in Microsoft Entra to update the callback url.
Another challenge which I had faced was setting up slack workspace and channel and adding in the oauth permissions. I had seen issues of messages not getting scanned due to proper permissions not being added and bot not being invited in the channel. Once done, the message scanning started working well.
Accomplishments that we're proud of
This project allowed me to create a new ops dashboard which can be used by engineering teams. By turning workplace conversations into structured Jira tickets, triage alerts, emails, Confluence documentation, and EOD summaries, it reduces repetitive administrative work and allows teams to focus on resolving problems.
In terms of technical accomplishments, this project allowed me to learn about website hosting on Render, configure Microsoft OAuth through Entra, and integrate multiple platforms including Slack, Jira, Confluence, Outlook, and OpenAI.
What we learned
This project allowed me to learn majorly about website hosting on Render and secure enterprise integrations. Apart from this, I also learned how to effectively use Codex and the GPT-5.6 Sol model. I have never worked with codex before so this project allowed me to learn about how efficiently the tool works and how it's shapes the project in the best way possible.
What's next for Ops Pilot
1.) Integrating Ops-Pilot with teams via microsoft enterprise version 2.) Adding proper monitoring of AI agent on Graffana 3.) Rolling the platform out to other engineering teams to test it out on their own slacks/teams channels and meets 4.) Using Microsoft Entra to use AD based easy provisioning/removal of users from the portal.
Log in or sign up for Devpost to join the conversation.