How well is your AI transformation steered?
Ten questions in two sections: the foundations of steering, then the four decisions, build, scale, automate and stop. Together they give you a starting point. This is Stage 0 of the AI steering cycle: know where you start.
- Indicative self-assessment, not a measurement. It reflects your own answers. It does not compare you with other organisations, and it does not measure your use cases.
- Your answers stay in your browser. Nothing is sent, nothing is stored, no cookie. Reloading the page clears them.
- Not sure is a valid answer. In the score it is left out and never counts as zero, and it becomes something to find out. Skipping a question counts as Not sure, except the confidence question (question 10), where skipping reads as Low confidence.
It takes a few minutes.
The ten questions
Your browser is not running scripts, so the result cannot be worked out on this page. All ten questions are below, and "How this is scored" explains every rule, so you can score your answers on paper.
Part 1, Foundations. Question 1 of 10
Part 1, Foundations
Question 1
Question 2
Map where AI really works
Question 4
Question 5
Question 6
Question 7
Part 2, The four decisions
Question 8
Now, what your organisation has actually done, how it decided, and how sure you are.
Question 9
Question 10
Your indicative result
Indicative, from your own answers. It is not a measurement and not a benchmark. Your answers stayed in your browser; nothing was sent.
A real picture comes from measuring every use case, not from a questionnaire.
Your indicative stage
Exploring, the first of four steering stagesOrganising, the second of four steering stagesSteering, the third of four steering stagesSustaining, the fourth of four steering stagesToo much is not known yet to suggest a stage.
AI is used, but not yet tied to goals or seen as a whole.
From your answers, people use AI, but the organisation cannot yet say what for, or where it works. The first move is to write a few goals as outcomes. Then map every use case before rolling out more.
The picture is forming. Goals, map or decisions are still partial.
From your answers, parts of the picture exist: some goals, a partial list, perhaps a steering group. The next move is to make them meet: one map, measured the same way everywhere, read against the goals, and a first real decision.
Goals, measures and a governing body are in place. The cycle is running.
From your answers, goals, a shared map and a governing body are in place, and decisions are taken. The work now is depth. Raise each use case to the level its work justifies, one level at a time, with proof. Capture value on the whole line, not at one station. Decide on evidence, including what to stop.
The cycle runs on evidence, and stopping is normal.
From your answers, the cycle runs on evidence, and stopping is normal. Keep the goals current. Keep value spread across teams rather than carried by a few. Revisit what freed time is used for.
From your answers, too much is not known yet. That is itself a useful answer. Your next steps start with finding out. Your profiles and gaps are still shown below.
These four stages are unrelated to the cycle's Stage 0 to 6.
This is a reading of how your organisation steers AI, from your own answers. It is not Quilane's maturity scale, which describes each use case by the human's role.
Why the stage is not higher
- Your goals are not yet written as outcomes. Without them, the other measures cannot be read against anything.
- Your goals are only partly written as outcomes that each level carries. The stage cannot go further until they are.
- One foundation is not known yet.
- Two foundations are not known yet.
- No build, scale, automate or stop decision has been taken yet. Steering starts with deciding.
- Sustaining needs decisions on evidence, including stopping.
Your foundations
Six foundations, in the order the method sets them.
Goals, at each level, with expected outcomes:
Not yet in placeStartedIn placeWell establishedNot known yet, of 3Evidence needed
Map where AI really works:
Not yet in placeStartedIn placeWell establishedNot known yet, of 3Evidence needed
The system people work in:
Not yet in placeStartedIn placeWell establishedNot known yet, of 3Evidence needed
People, level by level:
Not yet in placeStartedIn placeWell establishedNot known yet, of 3Evidence needed
Value against expected:
Not yet in placeStartedIn placeWell establishedNot known yet, of 3Evidence needed
A governing body that reviews and decides:
Not yet in placeStartedIn placeWell establishedNot known yet, of 3Evidence needed
Evidence needed means at least one answer here was Not sure. The foundation cannot be Well established until it is known.
Your four decisions
Build:
Not startedStartedRepeatableNot knownRepeatable, on activity signals
Scale:
Not startedStartedRepeatableNot knownRepeatable, on activity signals
Automate:
Not startedStartedRepeatableNot knownRepeatable, on activity signals
Stop:
Not startedStartedRepeatableNot knownRepeatable, on activity signals
- How decided
- Evidence-led. What weighed most was goals, value, reach or proven use.Mixed. Evidence weighed, and so did requests or activity counts.Activity-led. What weighed most was requests, counts, estimates or offers.No decisions yet.Not known.
- Confidence
- Low ({n} of 10)Low (not answered)Partial ({n} of 10)High ({n} of 10)
Evidence needed
The basis or the confidence behind your decisions is not yet clear enough to rely on.
Your strengths
- Your goals are written as outcomes, and each level knows what it carries.
- You can see every use case and how the work is done.
- Shared methods have owners, and people search before they build.
- People are guided by role and check output before it leaves.
- You compare value with what was expected, beyond time saved.
- Your governing body meets on a rhythm and decides.
No foundation is in place yet. Start with goals.
Your gaps and next steps
Three things to address first. Each step is something your organisation can do itself.
Where you answered Not sure, the step is to find out. Not knowing yet is a useful answer, not a failure.
Keeping the cycle running
Revisit the goals: confirm them or change them.
Check whether value is spread across teams or carried by a few.
Review what freed time was used for.
AI is not yet tied to a few written outcomes that each level carries.
Write a few outcomes AI should serve, in business terms, each with a baseline, a target date and an owner.
AI is not yet tied to a few written outcomes that each level carries.
Ask each department which goals it carries, and agree them in both directions.
AI is not yet tied to a few written outcomes that each level carries.
Find out whether written AI goals exist, who owns them, and whether departments know which ones they carry.
You cannot yet see every use case, or what the person still does in each.
List every use case by department before rolling out anything new.
You cannot yet see every use case, or what the person still does in each.
For each one, note what the person does and what the AI does, on four levels, from the AI assisting to the AI producing.
You cannot yet see every use case, or what the person still does in each.
Ask each department for its list of AI use cases, and compare what comes back.
The gap between the level each department has built and the level its people actually work at is not yet read.
For each department, put side by side the level of the use cases it has built and owns and the level its people actually work at. Bring the largest gaps to your governing body.
The gap between the level each department has built and the level its people actually work at is not yet read.
Where people work at a different level from what their department built, look at which use cases they actually use, including those built by other departments.
The level people actually work at is not yet known.
Ask a few departments for both levels: what they have built and own, and what their people actually use. Compare the answers.
Useful ways of working stay with individuals, or multiply without owners.
Give every method used by several people an owner.
Useful ways of working stay with individuals, or multiply without owners.
Retire methods that have no owner, instead of adding approval gates.
Useful ways of working stay with individuals, or multiply without owners.
Ask a few teams where they look before building something new, and who they tell when a shared method breaks.
Access came before guidance, and progress is not shown by results.
Have each team list its recurring tasks, and how to do each one with AI.
Access came before guidance, and progress is not shown by results.
Spread four checking reflexes before anything leaves: question, challenge, validate, judge.
Access came before guidance, and progress is not shown by results.
Ask a few teams what guidance they received after access, and how they check output before it leaves.
Value is inferred from usage, or from estimates of time saved.
Set an expected value for each use case before it goes live, then compare actual with expected.
Value is inferred from usage, or from estimates of time saved.
Before speeding up a team, ask who receives its output, and whether they can keep pace.
Value is inferred from usage, or from estimates of time saved.
For your largest use cases, find out what value was expected and what was observed.
Decisions are taken case by case, or the steering group mostly hears updates.
Fix a regular rhythm, and start each meeting from what moved since the last one.
Decisions are taken case by case, or the steering group mostly hears updates.
End each item with build, scale, automate or stop, with a reason and an owner.
Decisions are taken case by case, or the steering group mostly hears updates.
Look back at the last few meetings. How many items ended in a decision with a reason and an owner?
Nothing has been stopped on purpose.
Agree a stop rule in words, for example: capture stays persistently below expected value and adoption does not move, over a set number of councils. Put the use cases that meet it, or serve no goal, on the next agenda with stop as an option. Stopping is a decision, not a failure.
Decisions rest on activity or requests, not on evidence.
For each decision, write which goal it serves and what value was expected, before it is taken.
The evidence behind decisions would be hard to show.
Record each decision with its reason, its owner and the evidence used, where the council can find it again.
Proven manual work has not been considered for automation.
List the use cases that are frequent, proven and still manual, and review whether each should move to a level where the AI chains the steps.
Working use cases stay where they started.
Where a use case captures the value expected but reaches few of the people it should, decide whether to extend it.
New use cases appear without a decision.
Route new proposals to the governing body, grouped by business purpose, and attach each one to a goal.
From here
This is your starting point. The next step in the method is to set the goals it points to.
A real picture comes from measuring every use case. Quilane, at launch, measures each AI use case on maturity, adoption and capture across departments, and prepares each council from what moved since the last one. It shows department maturity in two readings, Built and Used, for every department and the whole organisation. Built is the average level of the live, adopted use cases a department owns. Used is the level of the use cases its people actually used in the last 30 days.
The Used reading is never shown when it rests on fewer than 5 people, to anyone, the council included. Employees who turn on "Hide my usage" are excluded from it.
A demo conversation on the sample workspace, oriented on your answers. Not consulting, not an audit, not a measurement.
Your results are copied. Nothing was sent.
Your browser did not allow copying. Use Print my results instead.
This clears every answer.
Quilane self-assessment. Indicative, from your own answers, {date}.
AI transformation self-assessment, {date} Indicative self-assessment, not a measurement. From my own answers. Not a benchmark. Indicative stage: {Stage name}, {stage one-line meaning} {caps} Foundations - Goals, at each level, with expected outcomes: {p0}{f0} - Map where AI really works: {p1}{f1} - The system people work in: {p2}{f2} - People, level by level: {p3}{f3} - Value against expected: {p4}{f4} - A governing body that reviews and decides: {p5}{f5} Four decisions - Build: {d0} - Scale: {d1} - Automate: {d2} - Stop: {d3} - How decided: {hd} - Confidence: {cf}{f6} Gaps and next steps 1. {g0l} Next step: {g0s} 2. {g1l} Next step: {g1s} 3. {g2l} Next step: {g2s} Method: https://{site}/methodHow this is scored
Everything is worked out in your browser, with the rules below and no others. You can score your answers on paper and get the same result. The thresholds are a design choice of this method. They are not calibrated against data from other organisations.
Each question gives a score from 0 to 3, or "unknown"
- Questions 1, 5 and 6 (one line to pick). The line you pick scores 0, 1, 2 or 3, from top to bottom. Not sure or skipped is unknown.
- Questions 2 and 3a (checklists of five). Three times the share of items ticked. For example, three items ticked scores 1.8. "None of these" scores 0. Not sure or skipped is unknown.
- Question 3b (one line to pick). We know neither: 0. We know one of the two, either one: 1. Both, for the organisation as a whole: 2. Both for each department, with the gap read: 3. Not sure or skipped is unknown.
- Questions 4 and 7 (yes, no or not sure per line). Three times the share of Yes among the lines answered Yes or No. Not sure lines are left out. If fewer than half the lines are answered Yes or No, the question is unknown.
- Question 8 (the four decisions). Not scored as a number. Each decision gets a state: Not started, Started, Repeatable, or Not known. Started and Repeatable count as practised.
- Question 9 (up to two picks). Four picks count as evidence: the business goal it serves; actual value against expected value; how far it reaches the people or cases it should reach; how often it is used, and whether it is proven and still manual. Four count as activity or opinion: a sponsor or budget holder asked for it; the number of licences or active users; people's estimates of time saved; what a vendor offered. All picks evidence: Evidence-led. Half: Mixed. None: Activity-led. "No decisions taken yet": No decisions yet. Not sure or skipped: Not known.
- Question 10 (the slider). 1 to 4 is Low, 5 to 7 is Partial, 8 to 10 is High. Skipped is Low: not knowing how sure you are is itself low confidence. The slider never raises a score. It can only add "evidence needed".
Each foundation
- Goals is the average of questions 1 and 2. If one of them is unknown, the other is used alone.
- Map is the average of questions 3a and 3b. If one of them is unknown, the other is used alone, and Map carries "evidence needed". Map is not known yet only when both are unknown.
- System is question 4. People is question 5. Value is question 6. Governing body is question 7.
- Evidence needed. A foundation with any Not sure answer in it carries "evidence needed", and its band can go no higher than In place, so it is never Well established.
- Scores are shown with at most one decimal. Every comparison uses the exact score.
Not sure counts as a gap, not as a zero
- In the score, a Not sure answer is left out. It never counts as zero, but it can limit the band and the stage, and it becomes a step to find out.
- Skipping a question counts as Not sure. The one exception is question 10, where skipping reads as Low confidence.
- In the result, it shows as Not known yet or as evidence needed.
- It limits how high the stage can go. Not sure can never raise a stage.
- When gaps are ranked, a foundation that is not known yet is placed after one that is clearly missing and before one that is started.
The four decisions
Each decision shows its state from question 8, then two lines for all four together: how decided (question 9) and confidence (question 10). "Evidence needed" appears when confidence is Low, when the basis is Activity-led or Not known, or when a decision is Repeatable while the basis is Activity-led. That decision then reads "Repeatable, on activity signals".
The steering stage
- Overall score. The average of the known foundation scores. Each foundation weighs the same.
- Base stage. Below 1: Exploring. From 1 to below 1.75: Organising. From 1.75 to below 2.5: Steering. From 2.5: Sustaining.
- Limits. Every limit that applies is used, and the lowest stage wins. Each one is shown with your result.
- Goals below 1, or not known yet: Organising at most.
- Goals below 2: Steering at most.
- One foundation not known yet: Steering at most.
- Two foundations not known yet: Organising at most.
- Three or more foundations not known yet: no stage is shown. Your profiles and gaps still are.
- No decision practised in question 8: Organising at most.
- Stop not practised, or the basis Activity-led or Not known, or confidence Low: Steering at most.
- Nothing else. No weighting by sector or size, no comparison with other organisations, no hidden rule.
Goals limit the stage first. Without written goals, the other measures cannot be read against anything.
Strengths and gaps
- Strengths. Up to two foundations scoring 2 or more with no "evidence needed", highest first. On a tie, the earlier foundation in the list comes first.
- Gaps. Up to three. First, foundations below 2 or not known yet, lowest first. Goals always comes first when it is a gap. Then, if fewer than three, decision gaps in this order: Stop not practised; basis Activity-led or Not known; confidence Low; Automate not practised; Scale not practised; Build not practised.
- Steps. One per gap. A gap that is not known yet, or needs evidence, gets a step to find out. Otherwise it gets the first step for that gap. Map has two versions. If the only unknown Map answer is 3b, the step to find out asks about built and used levels. If 3a is unknown, it is the standard Map step to find out. If both Map answers are known and 3b scores lower than 3a, the gap uses the built and used line and steps. In every other case, it uses the standard Map line and steps. Three steps are always shown: with fewer than three gaps, the next step of the first gap is added. With no gap at all, you see three steps for keeping the cycle running.
What this is not
- Not a measurement, a diagnosis, a certification or a benchmark.It reads your own answers. It does not compare you with any other organisation.
- Not Quilane's maturity levels.Quilane describes each use case from L1 Manual to L4 Autonomous, by the human's role. This page gives a reading of how your organisation steers, from your own answers. A higher maturity level is not the aim for every use case.
- Not the map.The detailed, use-case-by-use-case picture is Stage 2 of the cycle: map where AI really works.
- Not your goals.The baseline tells you what the goals must address. Setting them is the next step.
- Not a prediction.No stage predicts a return, a saving or any business result.
- Not a Quilane feature, and not advice.It is a free page on this site. Quilane does not assess, audit or advise organisations. Leadership decides, through its council.
Book a maturity review
A maturity review is a demo conversation. We walk through the steering cycle on the sample workspace, oriented on your answers on the six foundations and four decisions. It is not consulting, not an audit and not a measurement of your organisation. Your answers on this page are not sent with the request.
A demo conversation on the sample workspace, oriented on your answers. Not consulting, not an audit, not a measurement.