My six-year-old is at the piano. He’s been at it three months. He’s playing “Hot Cross Buns,” and he isn’t thinking C, D, E. He’s thinking thumb, first finger, second finger. The notes have numbers, not names. Three notes. On repeat. The song has three notes.
Nobody in the house thinks he should be playing Mozart. Nobody is disappointed. Nobody is asking when he’s going to compose. He’s six. He’s learning the instrument. The job right now is the instrument, not the music.
Then he closes the piano. The same parent walks into a conference room. The Tuesday standup dashboard is up on the screen. Ops manager built it in her spare time. She’s been using AI for ninety days. The dashboard works. The numbers are right. It also looks like a kid practicing scales.
Operator looks at it. Says it looks janky. Spikes the project. Team’s back on the spreadsheet from 2014 by Friday.
That move kills more AI deployments than any technology problem ever has. Not because the dashboard was bad. Because the audience was wrong.
The category mistake
Internal work and external work are different categories.
Tuesday standup is not the board meeting is not the customer-facing artifact. Different audiences. Different stakes. Different tolerances for mechanical-sounding output. Most operators treat them the same. Same polish bar. Same Mozart standard. Across all three.
That’s the lever.
The board deck is external in posture even when nobody outside the building sees it. The owner is performing competence for his own peer group. He’s been doing it for twenty years. The instinct to polish is wired in. So when the ops manager hands him the dashboard the team uses to run Tuesday standup, he reads it with the board-deck eye. He sees mechanical. He concludes broken.
Here’s the thing. The dashboard isn’t for him.
The dashboard is for the dispatcher who needs to see whether four trucks are out before she answers the next call. It’s for the foreman who needs to know which of his three crews is behind on Wednesday. It’s for the office manager who has been juggling the same five spreadsheets and just got handed one that automatically tells her what she used to spend forty minutes assembling. None of those people need a board-deck polish. They need accurate. They need fast. They need useful enough. That’s it.
The owner showing up and saying “this looks janky” is the same move as walking into the living room and telling a six-year-old his Hot Cross Buns sounds mechanical. Yeah. It does. He’s six.
Output Over Process is the ingredient most operators ship blank
The Professional Recipe has seven ingredients. Every AI station needs all seven to hold up. Training. Context. Guardrails. Examples. Output Over Process. Measurement. Feedback Loop.
Ingredient #5 is Output Over Process. Define what useful enough looks like before you define what polished enough looks like. The Chef sets the bar. The bar has two settings. Useful enough is the floor. Polished enough is the ceiling. Most operators conflate them. They set one bar, they set it at the ceiling, and they apply that bar to every artifact the team produces.
That’s the failure.
The Tuesday standup dashboard has one job. Help the team run Tuesday standup. Useful enough means the numbers are right and the screen is readable. That’s it. If the colors are mismatched, if the layout is awkward, if the header font is the default, it doesn’t matter. The team can run the standup. The artifact did its job. Polish is a separate decision the Chef makes later, when he decides whether the artifact has earned the time it would cost to polish it.
Owner who spikes the project on day ninety hasn’t done that math. He’s looked at the artifact, applied the ceiling bar by reflex, found it short, and concluded the tool doesn’t work.
The tool works. He’s grading the wrong test.
Internal artifact, internal audience, internal bar. The artifact passed.
The grace gap
There’s a parallel here I keep coming back to. Most operators have a remarkable amount of grace for people learning.
Same operator who spikes the dashboard would never tell the new hire, on week three, that her customer call notes look mechanical. He’d say “you’re getting there, keep going, here’s what to listen for.” He understands she’s learning the instrument. He gave her runway. He didn’t expect Mozart at week three.
He gave the AI dashboard ninety days and one read.
The grace gap is the whole problem. The team is learning the instrument. The artifact is the instrument the team is learning on. The owner who would protect a new hire’s runway while she got good at calls won’t protect the team’s runway while they get good at AI. He’s holding the AI artifact to a Mozart standard he wouldn’t hold anyone in his building to.
I mean. He spent a decade learning how to read a P&L. The dashboard’s been alive ninety days. Right?
The risk we have to name
The pre-empt on this argument matters. Hot Cross Buns is played correctly. The notes are right. Polish is the only thing missing.
Internal work has the same standard. Accuracy is non-negotiable. The dashboard better show the right four-truck count. The recap template better summarize the right meeting. The prep doc better be working from the right inputs. If accuracy is wrong, that’s a real problem, and the Chef should pull the artifact.
This piece isn’t permission to be sloppy. It’s permission to let the team play correct-but-mechanical for ninety days while they get the instrument under their hands. Different category of failure. Wrong notes is a deployment problem. Mechanical sound is a polish problem disguised as a deployment problem. Most operators are spiking on the second one and calling it the first.
The Chef’s job is to know which is which. Right? If the artifact is accurate, the team uses it. If the artifact has six fixes that would take an afternoon and earn the team another six months of compounding speed, the Chef calls for the afternoon. If the artifact is fundamentally broken, the Chef pulls it. None of those moves is “spike the project because it looks janky.”
That move is a different move entirely. It’s a Hub-level regression. The Chef who spikes the dashboard collapses from DIALING (taste and adjust) back to DOING (build it from scratch in the spreadsheet again). The Four D’s name this. The dish was at DIALING. The Chef tasted it. Instead of adjusting the recipe, he spiked the dish and went back into the kitchen. Same dish gets cooked from scratch every Tuesday for the next decade. Yeah. That’s the failure.
The team is your audience
The audience for internal work is the team, not the board.
Say that out loud. The audience for internal work is the team, not the board.
The Tuesday standup dashboard is for the team. The Slack recap your ops manager is generating with AI is for the team. The prep doc the dispatcher is using to start her shift is for the team. The first draft of anything internal is built for the team that’s going to use it, not for the operator who’s going to read it through his board-deck eye.
When the operator shows up to that artifact, his job isn’t grade it. His job is ask the team if it’s useful. If yes, the dish passes. If no, what would make it useful is the next question. Not what would make it look polished.
The team will tell you what they need. They’ll never tell you “I need it to look less mechanical.” They’ll tell you “I need it to also show open tickets,” or “the truck count should refresh every fifteen minutes,” or “can it ping me when crew three is behind.” Those are the notes the Chef pays attention to. Those are the adjustments that get the dish from useful enough to the thing the team can’t run a Tuesday without.
Polish, when it shows up, shows up because the artifact earned it. Not because the owner demanded it on day ninety.
The Monday Move
Pick one internal artifact your team has been building with AI. A dashboard. A recap template. A prep doc. The thing that’s been alive for thirty, sixty, ninety days. The one you’ve been quietly side-eyeing.
Look at it honestly.
Ask: is it accurate? If yes, ship it to the team this week, exactly as it is. Don’t fix the colors. Don’t change the layout. Don’t ask the ops manager for a polish pass. Just send it.
Then tell the team out loud, in the words that match how you actually talk: “This is Hot Cross Buns. We’re learning the instrument. We will not polish this until it earns the polish. If anything is wrong, tell me. If anything is missing, tell me. Everything else, we live with for ninety more days.”
Watch what happens.
Most teams find the artifact gets used and gets better. Not because anyone polished it. Because it stopped being precious. People started telling you what they actually needed instead of waiting to see if the next version would look better than this one. The polish that eventually shows up will be earned. The team will know what they want it to look like, because they’ll have lived with the mechanical version long enough to know which parts of it matter.
The instrument got under their hands. The music came after.
So.
The audience for internal work is the team, not the board.
My son will play Mozart someday. He’s not going to get there by skipping Hot Cross Buns. Neither is your team. The Tuesday dashboard will be sharp in a year. It will look like nothing right now. The two facts are the same fact, and the one you’re paying attention to is the wrong one.
Drop the board-deck eye. Pick up the piano-bench eye. Watch the room change.
Original framework. Distilled from client work.
Framework spine: The Professional Recipe, Ingredient #5: Output Over Process. Read the full framework. Sister piece: Speed Beats Polish, the external twin of this argument.
~ source material · Original framework. Distilled from client work.
