AI Workflows · 5-minute read
What should you actually give an AI assistant to do? Made with Dots offers a useful starting point: an independent gallery displaying 853 examples when checked on 10 October 2026. It collects public projects, experiments and accounts of everyday work. The interesting question is which ideas could fit your own life or business.
OpenAI describes dots as ongoing assistants with a cloud computer and access to connected apps. The gallery brings that idea down to recognisable tasks: untangling schedules, preparing project work and helping make things.
Which Dots examples are most popular?
The clearest available measure is the gallery’s Most viewed leaderboard, which ranks entries by views on their original posts. These were its top five displayed entries on 10 October 2026:
1. Everyday coordination and first-pass review: 57,466 views. Antonio Leiva’s Keet example describes finding parking around a tight schedule and preparing an initial review of a student’s assignment. Leiva handled the parking payment himself.
2. Finding overlooked commitments: 21,690 views. Maximiliano Firtman’s inbox and conversation review surfaced an outstanding reply, a scheduling clash and a potentially suspicious login.
3. Email, Slack and calendar coordination: 18,063 views. Dan Shipper’s Boo account describes filtering messages and spotting a conflict between a proposed flight change and a meeting request.
4. Moving an idea towards launch: 14,685 views. Peter Yang reported app-launch preparation and a potential revenue opportunity. An identified opportunity should not be read as realised income.
5. Responding to travel disruption: 12,118 views. Paul Hudson’s train story describes checking ticket validity and live delays to find an earlier service he could catch. The account describes using his existing ticket.
These figures measure attention to particular posts. They cannot establish how many people use each workflow, its success rate or the most common task across all Dots users. The gallery says its rankings cover approved, visible entries, and counts can change. Its categories are browsing labels, not an adoption survey.
Four practical ideas to take from the gallery
1. Bring scattered commitments into one view
The first three ranked stories suggest a useful starting point: information already exists, but it is spread across messages and schedules. For your own trial, ask for a short briefing showing deadlines, unanswered requests and conflicts, with a source beside every item. Keep the initial assignment focused on finding and explaining what needs attention.
2. Give a project a clear next deliverable
A separate Chase AI demonstration shows a lead-intake dashboard built with sample data and deployed after approval. The post is labelled as a ChatGPT partnership. Treat it as a demonstration to investigate. A sensible first business experiment is a prototype with dummy records, a defined acceptance test and a person responsible for approving release.
3. Investigate everyday friction
The parking and train examples suggest another direction: give your assistant a specific constraint to resolve. Ask it to compare options, show the relevant conditions and explain uncertainties. For travel, that might mean preserving an arrival deadline. For an appointment, it could mean allowing enough preparation and travel time. Check live details before acting.
4. Keep creative direction in the loop
In a video-production account, @yulhee___ describes turning a call recording into an edited video while correcting problems through feedback. That is a useful model for creators: provide reference material, request a first cut and review the actual output. Check pacing, captions, sound, source rights and whether the work expresses the intended idea.
A useful first brief
Choose one task you already understand well enough to judge. Define the inputs, the output, the deadline and the decisions that require your approval. Connect only the information needed for that task, and check which tools are available in your account.
For example: Review these three project documents and identify the five most important unfinished actions. Link each action to its source, distinguish confirmed deadlines from assumptions, and prepare a one-page plan for my review. Ask me before sending messages, changing shared files or making purchases.
This gives the assistant a finish line and gives you something concrete to evaluate. If it gets the priorities wrong, explain the correction using an example. Expand the responsibility only when the results justify it.
Measure the work, including the checking
Record how long the task normally takes, how much review the assisted version needs and which errors matter. Shipper’s early account also describes permission and message problems, a reminder that impressive examples can coexist with friction. OpenAI itself advises reviewing consequential work.
For creators and small teams, the opportunity is more room to develop an idea and fewer loose ends competing for attention. Browse the examples, choose one recurring problem and run a small, reviewable experiment. The most valuable use case will be the one that earns a place in your actual working day.
Sources and method: This article uses the gallery’s displayed total, leaderboard and linked entry accounts, checked on 10 October 2026. Social-platform analytics and reported outcomes were not independently audited. The practical trial suggestions are editorial recommendations, not results from AGENTICA IX product testing.
Related: OpenAI DevDay 2026: 10 AI Features That Matter for Creators and Small Teams.