178 ideas that shaped a program
Everyone wanted AI, and nobody knew where to put their idea. So I built a front door. What came through it ended up deciding what I built, what I taught, and how my job worked.
By the numbers
The front door I built
Before, there was a form and six responses. I replaced it with a conversation, and the ideas started coming. I designed and built the whole system myself, including the Power Automate logic.
Ideas can close at any stage. Even a no gets a plain-language explanation and an alternative.
I expected tickets. I got stories.
I expected one-liners like “can AI summarize this?” What came in averaged about 400 words: the problem, the data, the outcome people hoped for, and experiments they had already run on their own. Nobody asked them to write that much.
These weren't tickets. They were ideas people were trusting me with, and often they didn't need permission to try. They needed help scaling what already worked.
Three buckets for every idea
Every idea landed in one of three places. That pattern became part of the AI training, so people now sort their own ideas before they submit. Most of the time it's the first bucket, and they just go build it.
Build it
Do it yourself today with a prompt, a template, or a GPT and the tools you already have.
Ask for help
Scope it together: sample inputs, what good looks like, and starter prompts to try first.
Pass it to an engineer
It needs connectors, data access, or a security review. Hand it off with a clear brief.
The board became my needs analysis
In ADDIE, analysis is usually a phase you finish and move on from. The intake board turned it into a continuous feed, written by the learners themselves. Every submission is performance-gap data: the task, the pain, the context, and the outcome they hoped for.
My training roadmap came out of the board, not a survey. It decided which guides, courses, and skills I built, and in what order. Because I built every answer to be reusable, later requests often matched earlier ones, and four to six months in I was adapting finished materials instead of starting from scratch.
Ticket to close, or idea to steward
| Aspect | Ticket to close | Idea to steward |
|---|---|---|
| Goal | Clear the queue | Grow the idea |
| First reply | Auto-acknowledgment | A human who read it |
| Output | A one-off answer | A reusable asset |
| Credit | Invisible | The submitter's idea, named |
| Success | Ticket closed | They bring you the next idea |
Read it like a letter
Reply as a person. Reflect back what you heard.
Ask what it's for
Find the real outcome under the request.
Build the thing
A guide, skill, prompt, or workflow. Whatever fits.
Make it reusable
Write it so the next person can use it too.
Hand it back with credit
It's their idea. Say so.
What broke, and how I fixed it
Flooded on day one
People had been waiting to submit their ideas.
FixBring stakeholders into prioritization. Every idea matters, but not every idea needs us to build it.
“Did it go through?”
People weren't sure they'd submitted anything.
FixPlay back the full brief, confirm, then submit, plus a thank-you email.
AI ranking didn't work
The first version scored and auto-sorted every idea, and it ranked most of them about the same.
FixHumans decide. Time saved is context, not the verdict.
Every idea someone submits is them trusting you with a piece of their work. Honor that, and adoption follows.