# The dots demo, take two | OpenAI DevDay 2026

## Executive summary

The video demonstrates 'dots,' an advanced AI agent designed to automate complex, multi-step real-world tasks. The agent's capabilities span travel planning (booking flights and hotels based on policy constraints), synthesizing unstructured user feedback, and managing communication loops by drafting responses and linking to development artifacts (PRs). The demo highlights the shift from manual, multi-stage operational workflows to automated, proactive agent execution.

## Key takeaways

- Automated Travel Planning: The agent successfully handles complex, constrained booking requests, including finding flights that are 'in policy' and 'refundable,' and locating suitable, geographically relevant hotel options (e.g., Silver Lake Pool and Inn).
- User Feedback Analysis Workflow: The agent can analyze large volumes of user feedback (e.g., from a user feedback channel) to identify major themes, pinpoint pressing issues, and determine gaps where solutions are lacking.
- Closing the Development Loop: The agent can manage the entire issue lifecycle, from identifying a gap to drafting follow-up communications in Slack, linking directly to relevant Pull Requests (PRs) to prompt users to retest fixes.

## Technical details

- User Feedback Pipeline: The current manual workflow involves reading channel reports, populating a master list, classifying items, and linking issues to verified linear tickets. The goal is to automate the identification of 'no fix themes' and close the loop by linking PRs to user communications.
- Agent Reliability and Visibility: The agent itself identified that the largest current issue is reliability—specifically, when the agent appears active but fails to reply or access necessary tools. Users struggle with visibility into the agent's actual progress.
- Constraint-Based Booking: The agent demonstrated the ability to apply multiple constraints simultaneously, such as requiring flights to be 'in policy,' 'refundable,' and optimizing arrival times to meet a specific meeting window.

## Practical implications

- Automating manual, multi-step operational processes (e.g., travel booking, data synthesis).
- Improving product visibility by providing clear status updates on agent actions.
- Streamlining the Dev/Ops feedback loop by automatically linking user feedback to development fixes (PRs).

## Topics

AI Agents, Workflow Automation, DevOps, Product Management, Generative AI, dots, Slack, Google Maps, Pull Requests (PRs)

Source: https://www.youtube.com/watch?v=fHEIw5CcN5U