Today vs. Ten Years

The VP of operations pulled up the AI inventory last Tuesday.

Twelve tools in use across the business. Six workflows running on agents. Three pilots in flight. The model bill was up forty percent quarter over quarter, and the wins doc had twenty-two entries on it, all real.

She walked me through the doc, line by line. “Marcus saves an hour a day on inbox triage. Quote turnaround went from a day to three hours. The Friday forecast summary writes itself now.” Every entry was a real win. The team had earned it. The bill was a real bill, every dollar traceable to a workflow that was actually doing the work.

And she could not tell me which way the building was supposed to lean.

That’s the gap I want to talk about. The wins were real. The motion was real. The lean wasn’t there. She had a left column. She had never written down the right one. And it’s the absence of the right column that makes a fast-moving AI program feel, six months in, like a treadmill at a forty-percent-higher cost.

The two columns

Here’s the thing. Most operators are working what I’d call the left column of AI.

The left column is the list of things AI helps you do faster, better, cheaper this week. Inbox summaries. Quote drafts. Renewal letters. Meeting recaps. Three rounds of revisions on the proposal in twelve minutes instead of two hours. The productivity layer. The browser-tab AI. The wins you can post on a shared doc on Friday.

The right column is a different list. It’s the list of what work will look like when AI is the operating layer instead of the productivity layer. When the workflow was redesigned around what AI does well, not bolted onto a workflow that was built for a human in 2014. When an assistant runs the inbox and the operator reads summaries. When the questions get sharper because the work below them is being handled by a station that runs whether the operator is at their desk or not.

Most operators have never written the right column down.

That’s the indictment. Not that the left column is wrong. The left column is real. The wins on it are earned. The bill traces to actual delivered work. The indictment is that without the right column on the wall, the operator can’t tell which left-column moves to stop making, because they don’t know where the building is supposed to lean.

You can’t make the move you can’t see.

The table

I have been carrying this table around in client meetings for the last few months. Ten rows. Left side, today. Right side, ten years out.

The argument is not a forecast. The argument is structural. The shape of the gap is real. The specific rows are debatable, and you can swap rows on the bench for rows on the wall as your business calls for. The shape of the gap stays.

Read it like a yardstick, not a roadmap. Each row is a question, not a verdict. The question is: where am I on this row right now, and what would the next move toward the right column look like in my business this quarter?

Some rows you’ll read and notice you’re already partway across. Good. That’s the row your business is leading on. Some rows you’ll read and realize you haven’t even started the climb. That’s not a failure. That’s just the row you didn’t know was on the wall.

Why the bottom is heavier than the top

Here’s the thing. The rows are not equally weighted.

The top of the table, email and individual tasks and the productivity-layer wins, that’s the surface. Visible, immediate, easy to celebrate. The bottom of the table, the operating layer and the org chart and the role accountability, that’s the architecture. Invisible, slow, hard to point at. Most operators are spending most of their AI energy on the top because the wins there are visible and they compound on a shared doc that people can scroll on Friday afternoon and feel good about.

The leverage is in the bottom. The wins there compound in the business.

This is where the Station Plan starts doing the work. The Station Plan is the radial floor plan for an AI-native business. The Chef at the hub making the call. The stations on the line doing the labor. The orchestrator routing work between them. The right column of the table is what Station-Plan-shaped operating looks like at scale. The left column is pre-Station-Plan behavior, where AI is bolted onto the workflow you already had as a productivity layer, instead of integrated as a station on the line that owns a function and runs the function whether you’re in the building or not.

The VP with the wins doc had AI in twelve browser tabs. She did not have AI on the org chart. Nobody on her team had the role “runs the AI’s work.” Nobody had the role “decides which dishes get cooked by which station.” When she looked at her AI program, she literally saw the dishes, not the kitchen.

Twelve dishes plated. No kitchen.

Right? That’s the diagnosis the table makes visible. Without the right column on the wall, you can’t see that the dishes are happening without a kitchen, because the dishes are the only thing the doc is tracking. The kitchen never gets a row.

Why the right column is hard to draw

Most operators can describe their left column in five minutes. They can rattle off the tools, the workflows, the wins, the bill. The right column takes them a week. Sometimes a month. Sometimes they never get there.

It is not because the right column is mystical. It is because the right column requires the operator to imagine their own business with the work redesigned around what AI does well. That redesign is uncomfortable. It implies that some of the work the operator has been doing for a decade was the wrong work. It implies the org chart will change. It implies that next year’s job description for the office manager looks different from this year’s, in a way that nobody on the team has volunteered to name.

So the right column stays blank. The operator keeps banking left-column wins. The bill keeps climbing. The lean never arrives.

I want to be careful here. I am not saying the operator is wrong to work the left column. The left column is where you build the trust that AI is going to work in the first place. You can’t skip it. You can’t draw the right column from a standing start. You earn the right to read the right column by surviving the left.

What I am saying is that there’s a moment, usually six to twelve months in, where the left column has done its job. Trust is built. The team is not afraid of AI anymore. The wins are real. And the operator who never lifts their head and starts drawing the right column is the operator whose AI investment turns into motion without direction.

Twelve tools. Six workflows. Three pilots. Bill up forty percent. Wins doc with twenty-two entries on it. And no one on the team can tell you which way the building is supposed to lean.

That’s the cost. The first peak of Mount Stupid was the first AI win. The second peak is the right column. The discipline of looking up is the prerequisite for being able to read it.

The bottom row is the move

If you only have time to look at one row this week, look at the bottom one. AI is a productivity layer / AI is the operating layer. That’s the architectural pivot. Every other row in the table is a symptom of which side of that one you’re on.

A productivity-layer operator buys tools. An operating-layer operator builds stations. A productivity-layer operator measures outputs. An operating-layer operator measures functions. A productivity-layer operator asks “did the AI work today?” An operating-layer operator asks “did the role work this quarter?” All ten rows fall out of the bottom row. If you flip the bottom row, you flip the table.

That is the move worth picking this quarter. Not all ten. The one that decides the other nine.

The Monday move

Print the table. Hand a copy to your second-in-command. Sit down for twenty minutes. Go row by row. For each row, mark where the business is right now with an X. Not where you wish it were. Where it actually is.

Then pick one row, just one, where you are solidly in the left column and the right column is the row that pulls you. Write down one move you would make this quarter to step toward the right side of that row. Not all ten. One. The right one.

Does that make sense? You don’t have to make all the moves. You just have to know which way the building is supposed to lean. The X-marks on the table are the diagnosis. The one move you pick is the prescription. Everything else stays on the wall as the lean you’re building toward.

Plant the flag. Then look up. Then draw the right column.

You can’t make the move you can’t see.

Original framework. Distilled from client work.

Framework spine: The Station Plan, the architectural shift from AI as productivity layer to AI as operating layer. Read the full framework. Series so far: Mount Stupid (the discipline of looking up) and Five Questions for the Blind-Spot Prompt Ritual (the seventh ingredient).