Tech • AI • Robotics • Game
The dots demo, take two | OpenAI DevDay 2026
9.2/10TL;DR
A launch demonstration highlighted an AI assistant called Dot handling business travel, user-feedback triage, and message drafting, while also surfacing reliability and mobile UX as the most urgent product issues.
KEY POINTS
Travel booking under company policy
Dot was asked to arrange a last-minute trip to Los Angeles for a meeting in Hollywood at 3 p.m., with the traveler unable to leave the office before 10 a.m. The request included an on-policy, refundable flight plus hotel options in East LA with screenshots and map views to help compare locations. The assistant initially surfaced a Southwest option leaving at 11:30 a.m. and landing at 12:55 p.m., with a refundable fare around $326, then reconsidered earlier arrivals to create more buffer before the meeting.
Real-time itinerary adjustments
The assistant was also asked to judge whether landing around 1 p.m. would still allow arrival in East Hollywood by 3 p.m. Based on expected airport exit time and traffic, it estimated 30 to 45 minutes to leave LAX and roughly 45 to 75 minutes for the drive. That led to a preference for a flight arriving closer to noon, showing the system being used not just for booking but for time-risk assessment.
Hotel selection narrowed quickly
Hotel options included Silver Lake Pool and Inn and Cara, both presented as front-runners based on location and fit. After reviewing images, the traveler selected Silver Lake Pool and Inn, and Dot moved to prepare a one-night booking while verifying full price and cancellation terms. The exchange illustrated a workflow in which the assistant gathers options, provides visual context, and then executes once a choice is made.
Manual feedback pipeline remains in place
Beyond travel, attention shifted to the company’s overloaded user-feedback channel. The current process was described as a manual batch workflow: messages and replies are reviewed, entered into a master list, classified into themes, linked to verified Linear tickets, and matched to pull requests. A dashboard summarizes the work, but it does not automatically capture new messages, and ticket reconciliation is still incomplete.
Focus on unresolved themes, not just reporting
The requested next step was to make the feedback workflow more concrete and action-oriented. The priority was identifying the biggest issue clusters that still lack a credible fix, rather than simply summarizing what users are saying. Another requested improvement was a tighter loop between engineering and users, so that when a pull request lands, users in Slack are notified with a link and asked to try the fix again.
Reliability emerged as the top concern
In the most recent 24 hours, the strongest pattern was described as reliability: cases where Dot appeared active but failed to respond or could not access tools. The issue was framed as partly technical and partly a visibility problem, because users often cannot tell whether the assistant is working, stalled, or blocked. The practical result is uncertainty around missing replies, tool failures, and progress updates.
Mobile and notification polish also stood out
Alongside reliability, the assistant identified UX polish needs, especially around mobile behavior and notifications. That suggests the immediate product challenge is not only core system dependability but also clearer signaling and smoother interaction across devices. The mention of these issues places user trust and usability alongside raw capability as central launch priorities.
Broader communications support was folded in
The assistant was also tasked with collecting top unanswered Slack messages related to go-to-market questions, including how Dots relate to products such as chat, Slack, and Teams integrations. It was asked to draft responses, turning the system into a communications aide as well as an operations tool. The demonstration ended with a request to rerun the feedback workflow, reinforcing the idea of an assistant that can cycle through ongoing work rather than complete only one-off tasks.
CONCLUSION
The demonstration showed Dot being used for practical operational work, from booking travel to organizing product feedback and drafting internal responses. It also underscored that the most urgent obstacle to wider trust is still reliability, followed closely by clearer mobile and notification experiences.
Ask a question