You can turn ChatGPT’s voice feature into a 10–15 minute morning meeting with yourself: set one clear project, talk out loud while it listens and asks questions, then end with a structured summary and next steps. This workflow was used on a daily dog walk to shape PREP, an internal web app for kitchen staff that organizes prep lists and communication across shifts.
The Project: PREP in One Paragraph
PREP is an internal web app built for a working kitchen. It turns messy handwritten prep lists and verbal handoffs into a structured mise en place system for each station. Staff tap items from a shared prep library to build the day’s list, track what is done, and flag what needs to become part of the next order. It started as an industry frustration—there is plenty of software for POS, reservations, and inventory, but very little that respects how real prep and shift handovers actually work.
The Core Idea: A Daily Voice Meeting in Your Pocket
Most mornings begin with a dog walk in the park. That 20–30 minute block became a standing product meeting for PREP. The pattern stayed simple:
- Pick one project, not five.
- Tell ChatGPT what role it should play in the session.
- Talk out loud about the project while it listens and asks clarifying questions.
- End with a summary of decisions, open questions, and next actions.
That structure matters more than the technology. The value comes from repeating the same short process often enough that the project keeps moving.
Step 1: Give the Session a Job
Before talking, define exactly what the next 10–15 minutes are for. For PREP, the voice session can begin like this:
Good morning. For the next 15 minutes, I want to work on PREP, my internal kitchen prep app.
Your job:
1) Listen while I describe how kitchens handle prep and shifts today.
2) Ask for clarification if I'm vague.
3) At the end, summarize our meeting into:
- Decisions we made or leaned toward,
- Open questions I still need to answer,
- 3-5 concrete actions I can take before our next session.This framing keeps the conversation from drifting into random chat. It also pushes the model to behave more like a meeting note-taker and thought partner than a generic chatbot.
Step 2: Talk in Service Language, Not Consultant Language
The productive part of voice chat is not sounding polished. It is describing reality in plain language:
- How prep lists are created on a busy Friday.
- What happens when day shift leaves unfinished items for night shift.
- How missing prep turns into last-minute ordering, waste, or 86’d dishes.
The best material usually comes from concrete examples. Instead of saying “improve communication flow,” it is more useful to say, “The pantry cook needs to know by 4 p.m. whether the afternoon shift finished the sauces.” That is the kind of detail a useful tool is built around.
Simple questions from the model can also help:
- Who writes the prep list now?
- Where do staff note low items?
- What needs to be visible to the next shift?
Those questions sound basic, but they force the problem to become specific enough to design around.
Step 3: End Every Session With a Structured Summary
The summary is the part that turns talking into progress. Every session should end with a prompt like this:
Please summarize this as our PREP strategy meeting:
- Decisions we made or leaned toward.
- Open questions I still need to answer.
- 3-5 specific actions I can take before our next session.A useful PREP summary might include:
- Decisions: Version 1 focuses on digital prep lists and shift handover, not full inventory.
- Open questions: How should low items be logged quickly during service without slowing cooks down?
- Next actions: Sketch one station screen, seed the prep library with common items, and define what a good handover looks like.
That output becomes the bridge to actual work. It can be pasted into notes, a task app, or a running product document.
Step 4: Use Other AI Tools to Extend the Work
Once the voice session ends, the summary can be reused in different tools for different purposes:
- Claude: Turn the summary into a more formal planning document—problem statement, users, feature list, and roadmap.
- Perplexity: Research similar products, market feasibility, and realistic constraints around kitchen prep and back-of-house tools.
In practice, each tool plays a different role:
- ChatGPT voice = live meeting and rough transcript
- Claude = planner and scribe
- Perplexity = outside reality check
The important point is not using every AI tool available. The point is that one voice session can produce output that is useful in the next layer of planning and research.
Step 5: Make It a Routine, Not a Novelty
One good session is interesting. Repeated sessions create momentum. PREP moved forward because the same pattern happened over many mornings:
- Same project for several weeks.
- Same rough time window.
- Same closing question asking for a summary and next steps.
That repetition helped separate what belonged in version 1 from what could wait. It also made progress easier to resume later, because each session ended with a written snapshot of what came next.
A Template to Reuse Tomorrow
Here is a copy-paste script for anyone who wants to test this as a morning routine:
Good morning. For the next 10-15 minutes, I want to work on one project: [PROJECT NAME].
Your job:
1) Listen while I describe how things work today and what I would like to change.
2) Ask questions if I'm vague or contradict myself.
3) At the end, summarize our meeting with:
- Decisions we made or leaned toward,
- Open questions I still need to answer,
- 3-5 specific actions I can take before our next session.
I'll start talking now.The goal is not to “use AI more.” The goal is to have one short, productive meeting each morning that moves a real project forward. For PREP, that process helped turn kitchen experience into a working internal tool rather than just another idea left sitting in the background.