The free check · in your browser

Thirty-six questions about what you have already attempted.

Every question asks about something that has happened, with a date on it. Answer six for a band in five minutes, or all thirty-six in about twenty.

Answer the first sixWhat the six areas are
Free · 36 questions · about twenty minutes · nothing leaves your browser
Three concentric white rings lying at an angle, one chrome sphere sweeping around the outermost with a glass trail.
A reading, dated today
01What it reads

It reads what you now attempt, carry and finish.

Six areas decide how much an organization can now set out to do and actually finish. This check reads all six, with six questions each. Every question asks about one thing that has already happened. You should be able to show it to somebody this afternoon. A start date. A name on a page. A written record. A timed result.

So the questions are about attempts, not intentions. What did you set out to do this quarter that you would have declined two years ago? How far did one person carry it alone? How much of what you know is written down well enough to hand part of it over? What actually runs, and how fast could you replace it? Who chose the work, and who answers for it? And did the second round cost less than the first?

The check does not ask what you plan to do next year. It does not ask how many licenses you bought. Paying for a tool is not the same as being able to do something new with it. It does not ask you to rate yourself on a scale.

There is a reason for that. In a randomized trial, sixteen experienced developers were timed against what they believed about their own speed. Belief and the clock disagreed by thirty-nine points. The trial is set out in full, with its design, its dates, and what its authors have since said about it. A reading built on what people believe about themselves is worth nothing. A reading built on things that happened is worth the twenty minutes.

What to do first. Open last quarter’s records before you start. Most of these questions are settled by a date.

A long white lever resting on a chrome fulcrum, its far end lifting a large white cube clear of the ground.
One person, and what they now lift
02How long, and what you get

Five minutes for a band. Twenty for a reading you can print.

The short path is six questions, one for each area. It gives you a band with a plain name and no number, and it takes about five minutes. All thirty-six take about twenty minutes with last quarter’s records open beside you. You can stop after six and come back.

What comes out of the full set

  • A reading from 0 to 100, dated today, with one bar for each of the six areas.
  • Your first three moves, taken from your three lowest answers, heavier areas first.
  • The questions where you had least to show, which are the first rows of your plan.
  • A page you can print and hand to whoever will own those rows.

Your answers are saved only in this browser, on your own computer, so you can stop and come back. Press Reset to erase them. Nothing is sent to us, and nothing about you is published anywhere.

The written version of this reading is the report. It runs eight to twelve pages and comes back in under an hour. What you claimed sits beside what you could show. It is $290. Until card payment opens, it comes with the amplification check at no charge.

What to do this week. Take the short path now. Come back for the other thirty with your records open.

A chrome sphere resting in the upturned palm of a large white sculptural hand.
A reading you can hold, dated today
03The questions

Six areas, six questions each, five answers apiece.

The five answers are levels, from 0 to 4. Level 0 means it has not happened. Level 4 means it has happened more than once and is ordinary here now. Every level describes a state you could point at, never an adjective. Pick the highest one you could prove this afternoon with the record in your hand.

One rule sits over all thirty-six. Nothing scores until it has happened once. A plan scores nothing. A license scores nothing. A pilot that ended in a slide deck scores nothing. A stated ability to change supplier, with no timed record behind it, scores nothing either.

If nobody in the building knows the answer, choose Level 0. Not knowing is itself a finding, and that question becomes the first row of your plan. One question in each area carries the short path, and is marked. Where a question says the board, read the owner, the parent company, or whoever your most senior person answers to.

Before you start, one limit worth knowing. This reading rests on your own answers, and nothing you enter here is checked against a record. So take it twice: once from memory, and once with the records open in front of you. The gap between those two readings is the most useful thing the check will tell you.

68Sample · illustrative
A sample result · made up, and worked through on Your Index

Amplified

  • Ambition74
  • What one person can finish62
  • Working knowledge71
  • Machine capacity84
  • Direction64
  • Does the second time cost less55

First moves: lift Direction, which counts twice, then make the work compound. Machine capacity earned 92 and prints 84, because no area may read more than twenty points above Direction.

You leave with a reading from 0 to 100 and a score for each of the six areas. You also get three moves and a page you can print, dated today.

Area 1 of 6

Ambition

What your organization now sets out to do, against what it would have set out to do before.

Six questions about what you have started, what you turned down, and what you priced again.

1. Have you started something this quarter that you would have declined two years ago? Short path

Look for a piece of work with a start date on it, not an idea somebody liked in a meeting.

2. Do you keep a written list of the projects you declined because trying was too expensive?

One page: what you turned down, the reason, and what it would have cost you to try then.

3. Is there one named person who owns how much the organization attempts?

One name in writing, with what they decide and what they report. Not a committee.

4. Have you written down what the organization will not build at any price?

A short page of refusals, signed. It is what lets you say yes quickly to everything else.

5. How often is the declined list priced again at today’s cost of trying?

Prices fall fast. A project priced two years ago carries a cost that no longer exists.

6. How long does an untried idea take to reach a first working version?

Look at the last three ideas somebody proposed. Count the days from the proposal to the first thing that ran.

Area 2 of 6

What one person can finish

How far one person can carry a piece of work alone, from the idea to the finished thing.

Six questions about what one person now carries, and what the machines carry beside them.

1. Has one person carried a piece of work from idea to finished thing on their own? Short path

One name, one finished thing. Help from machines counts. Help from a team does not.

2. For the work you finished last quarter, do you know how many people it used to take?

One line against each item: who carried it, how many it needed before, and how long each way took.

3. In your main pieces of work, do the steps pass between people differently from two years ago?

If the same steps pass between the same desks, the tools were added on top of the old design.

4. Have your people been using one of the best AI tools on real work for months, with that work saved?

A license and a one-hour session is not months of use. Look for saved work they can open again and reuse.

5. How many people here direct several machines at once, rather than doing the task themselves?

Directing means setting the task, reading what comes back, and deciding what happens next.

6. When somebody needs a skill your organization does not have, what happens?

Two years ago the answer was to hire it or buy it in. Look at the last three times it came up.

Area 3 of 6

Working knowledge

What your organization knows about its own work, written down well enough to hand part of it over.

Six questions about what is written down, and what still lives only in somebody’s head.

1. Do you keep real cases with known good answers, in files you own? Short path

Real work with the right answer beside it. It is how you learn whether a new model is better at your work.

2. Can you put a list of what your organization actually does on the table within the hour?

One row for each piece of work whose result drives a decision, with an owner and a review date.

3. Is the work written down as the actions people take, or as job titles?

A title cannot be handed to a machine or tested against one. An action can.

4. Have you handed real work to a model and kept a record of what came back?

A meeting about whether a machine can do something produces an opinion. Handing it the work produces a record.

5. Does each row say what a machine may be handed and what judgment stays with a person?

Two columns beside the work. Somebody has decided where the line sits, and has written it down.

6. Is the material your people produce with machines kept where you can use it again?

The instructions people wrote, the graded cases, the records of what came back. Keep them where a change of supplier cannot take them.

Area 4 of 6

Machine capacity

What actually does the work, and how easily you could change the AI underneath it.

Six questions about what runs, how much room it has, and how quickly it can be replaced.

1. Have you replaced the model under a live piece of work, and written down how long it took? Short path

Being able to change supplier is what protects an ambition. A stated ability, with no timed record, counts for nothing.

2. Where does a model’s name appear in your systems?

Ask whoever looks after your software. One settings file means you can change supplier in an afternoon; names written inside the software mean a rewrite. If you only use apps you bought ready-made, ask whether the AI engine behind them is a setting you can change.

3. Do you have software that runs a job from start to finish on its own?

Not the kind that waits to be asked at every step. Software that takes the task and carries it to the end.

4. How much room does each machine-run process have to act on its own, and who decided that?

The room a machine gets is a choice about your business, not a fact about the model. Somebody can widen it.

5. Is the rule that pulls a process back to human approval after an incident written down?

This is the brake that lets you widen the room in the first place. Write it before you need it.

6. Has a second supplier’s model passed your own tests on real work in the last quarter?

A named alternative with a dated score. Not a supplier named in a document, and not a benchmark result.

Area 5 of 6

Direction

Who chooses what gets attempted, and who answers for it afterwards.

Six questions about who may decide, who stands behind it, and how much that lets you run at once.

1. Does each machine-run process have one named person who can stop it, so more of them can run? Short path

One name against each process, and a person who can be asked why. Not a committee, and not a queue.

2. Does every process that moves money, or touches safety or the law, have one named person who can say yes?

One person who allowed it and can be asked why. Whoever a license or a rule already names comes first.

3. How much machine-run work do you let run at once, because anybody here can stop one?

One named way for anybody here to stop a running process, without asking first. That is what makes the number safe to raise.

4. Are the person who chooses the work, the person who checks it and the person who signs different people?

Three names on one row. It is what lets an approval be quick and still mean something.

5. When a machine produces work that reads well but is hollow, does somebody catch it and say why?

Fluent and wrong is the failure that gets through. One rejection with a written reason is what makes the rest trustworthy.

6. Do the people who check machine work have the training and the standing to disagree with it?

European law expects whoever checks an AI system to be trained for it, allowed to say no, and backed when they do. (EU AI Act, Article 26(2).)

Area 6 of 6

Does the second time cost less

Whether this quarter’s work makes next quarter’s work cheaper.

Six questions about what you keep, and whether the second round costs less than the first.

1. Has a second round of the same kind of work cost less than the first? Short path

Cheaper because you kept the cases, the instructions and the grades. Not because people tried harder.

2. Do you write down what you expect to happen, then score it when the answer arrives?

Six columns: what you expect, how likely, by when, what would prove it wrong, who said it, and the score.

3. When somebody disagrees with a decision, do their name and their reason stay on the page afterwards?

Look for somebody’s objection still written down, after the decision went the other way.

4. Do scored calls decide anything, and is it written down what they never decide?

They should decide how much room a machine gets. They should never decide anyone’s pay or grade.

5. Do junior people do real, bounded work with machines under somebody who signs for it?

This is how seniors are made now. Look for a supervisor’s signature on a junior’s finished piece of work.

6. Do you re-run this reading, your list of work and your declined list on a fixed timetable?

Ninety days is the interval. A dated second reading sitting beside the first is the proof.

Not counted in the reading

One question sits outside the score, and it is worth answering anyway.

For the few pieces of work that must keep running, is it written down what runs instead? That is the whole question. It is not one of the thirty-six, and it changes nothing in your reading.

It is left out because the reading measures what you can now set out to do, and this is not that. It is the plan that lets you commit your most ambitious work to a foundation somebody else owns. It protects the ambition. It does not measure it.

What to do. Mark the rows on your list of work where a supplier failure would stop the business. Against each one write four lines. Who notices, what carries the work meanwhile, how long the alternative takes to stand up, and who decides to invoke it. Who: whoever owns that list. How long: one afternoon. You end up with: a short list, each row naming those four things. Done when: the list fits on one page and the alternative has been timed once rather than guessed. The route sets it out in full, sized like the plan you already hold for power.

04What the reading means

Four bands, and each one names a state of capability.

None of the four is a warning and none is a grade. Each describes what somebody walking around your organization would find on the day you took the check.

0–44

Manual

The work is done much the way it was done before. Tools are present, often bought one at a time and never recorded. Nothing anybody sets out to do has changed.

45–59

Assisted

The existing work goes faster. People are quicker at their own tasks, and the handoffs between them sit where they always sat. You can point to time saved and to nothing new attempted.

60–79

Amplified

Work that used to need a team has shipped, carried by one named person. Something has come off the declined list. Every process that matters has a person who can stop it, which is why several run at once.

80–100

Expanding

The limit is what you dare to choose. Each quarter’s work makes the next cheaper. A change of supplier has been performed and timed, and forward calls are scored against what happened.

One ceiling sits above the arithmetic. No area may read more than twenty points above Direction, which covers who chooses the work and who answers for it. Capability nobody will answer for does not stay in use for long. The reading prints what an area earned beside what the ceiling allowed, so nothing is hidden from you.

Areas are weighted. Ambition and Direction carry a fifth each. They are the two ends of one decision: what to set out to do, and who will stand behind it. The other four carry fifteen percent each. How the reading is built sets out the arithmetic in full, and the six areas explains what each one measures.

The reading carries the date you took it, and it goes stale in plain view. Take it again in ninety days and set the two readings side by side. What moved between them is the only progress number worth reporting.

What to do with it. Read the band above yours. Find the sentence in it that is furthest from true today, and put one name against that sentence this week. The route is written in the order those sentences usually get fixed.

Take the short path now, and the other thirty when your records are open.

Answer the first sixAsk us a question