Turning group chat chaos into productive decisions through agentic AI.
ORGANIZATION
Self-initiated project
CATEGORY
Agentic AI Experiences, Productivity, Communication
DURATION
1 Day
TOOLS
Claude Code, Figma
PROBLEM
Ever found it hard to lock in group plans, coordinate timings, and decide what to order… only to end up in an endless back-and-forth?
💬 A quick chat with a few friends about this surfaced the following:
"It's been soo hard to make plans lately… Just the process of deciding a time and place for meet that works for the entire group causes so much burnout."
"We can never really decide quickly."
"I feel like spontaneity is lost because everyone is so busy these days so we need to plan much in advance."
SOLUTION
An on-demand AI moderator that summarizes chaotic group chats, matches preferences and availability, runs voting, and then completes the plan through group ordering or table reservation.
DESIGN DECISIONS & RATIONALE
Fast ideation meant leaning on systems thinking. Here’s the assumptions and decisions behind the concept.
Summonable intervention
Forcing an agent presence all the time can feel intrusive.
DESIGN DECISION
Make the agent opt-in, so the group uses it when stuck, not by default.
Conversation Summary before Suggestions
People don’t align because the shared context is fragmented across messages. Not everyone keeps up with the last decided items or availability updates.
DESIGN DECISION
A structured summary (“what’s agreed”, “open questions”, “constraints”) so the group can catch up before acting.
Voting as the decision accelerator
In group settings, discussion often never converges because there’s no “mechanism” to close the loop.
DESIGN DECISION
Early voting-end options, controlled by group admin in case the majority is reached.
Chat drawer accessible during decision phase
If users leave the chat context to vote/order/book, they lose social cues, confirmations, and ongoing negotiation—causing drop-off and reopening debates.
DESIGN DECISION
Use a persistent chat drawer while in voting to maintain context.
Branching after the vote: “Order in” vs “Reserve”
After voting/finalising the plan, there is a need for facilitation of the next step: Reservation or Group Ordering.
DESIGN DECISION
Early voting-end options, controlled by group admin in case the majority is reached.
POSSIBLE BREAKDOWNS
There are some areas in the flow where this concept has chances of breaking down.
FUTURE CONSIDERATIONS
Due to the limited sprint timeframe, the following areas were noted for deeper exploration in future iterations.
REFLECTION
Reflecting on AI usage and implications.
This concept was vibecoded with AI to move fast, but I treated it as an early systems prototype, not a validated solution. Since an agent that summarizes and nudges decisions can easily steer group outcomes, I designed for user control and transparency through summonable intervention, and clear closure states rather than relying on AI authority.
If I take this forward, my priority would be user-centered validation around fairness, consent, and edge cases to ensure it reduces friction without overriding minority constraints, exposing sensitive preferences, or creating false confidence in the plan.