Managers now decide how their teams take up AI. They build their own AI skills, choose where AI goes into the work, help people through what it means for their roles, and stay accountable for what AI-assisted work produces. Leading with AI measures how well they lead their teams’ use of AI across all of these. It serves candidates and current managers alike, from team leads to executives, and suits change and enablement teams too.
We built it in two parts. The first decides what to measure: which skills earn a place, and what evidence backs them. The second decides how to measure them. Each scenario tests one skill, follows the same set of rules and passes a set of checks before it’s used. The assessment is built on a job analysis and peer-reviewed research, confirmed by subject-matter experts and validated by I/O psychologists. It has 25 scenarios, five for each skill, each with four options, and takes about 15 minutes.
Part one: deciding what to measure
The evidence base
The skills rest on two kinds of evidence: a job analysis of the work managers now do as AI changes their teams’ work, and published research on how people and teams take up new technology and AI.
For the job analysis, we collected job descriptions for team leads, managers and senior leaders across functions and regions, and kept those with real people-management responsibility and responsibility for AI tools or AI-driven change. We coded each one for the leadership behaviors it requires, and kept the behaviors that appeared consistently across functions rather than in one role only.
We then checked the results against the research, which covers what helps a team adopt a new way of working and what makes AI adoption specifically go wrong, and against practitioner frameworks for leading AI adoption. Subject-matter experts in people management and AI adoption confirmed the final skills. Every claim the assessment makes about leading AI adoption rests on this evidence.
Strong performance and failure modes
For each part of a manager’s work that AI now touches, we set out two things.
Strong performance. What good leadership looks like when AI is part of the team’s work. When a team starts using AI for client reports, that means deciding which steps the AI takes on, and who checks the result before it goes out.
Failure modes. A failure mode is a specific way AI adoption goes wrong under a manager, whether the team moves too fast with AI or is held back from it. Each one records what goes wrong, why nobody notices, and what a capable manager without experience of leading AI adoption does instead.
The exclusion test
Every failure mode then faces one test: take AI out of the picture. If the problem still happens, it belongs to people management in general, and we exclude it. We keep only the failure modes that exist because AI is changing the work. That keeps the assessment measuring how a manager leads AI adoption, rather than management itself.
Five gates a skill has to pass
We group the failure modes into candidate skills. Every candidate skill has to pass the same five gates, in order. If it fails any one of them, it’s cut.
Gate | The test |
|---|---|
Specific to leading AI adoption | The problem shows up when a manager leads a team through AI adoption. If it would read the same for any individual using AI, it belongs in the general AI Fluency Assessment instead. |
Not already covered | The general AI Fluency Assessment doesn’t already measure it, or doesn’t measure it at this depth. |
Measurable in a work scenario | A single leadership scenario can measure it, with a graded score from the best response to the weakest. |
Turns on AI, not on AI knowledge | Replace AI with any new software system and the best action changes. Yet a capable manager can reach the best action from the facts of the scenario, without knowing how AI tools work. If the answer depends on technical knowledge of AI, the skill belongs in the AI Fluency Assessment. |
Enough evidence to measure | The skill can be measured through several different scenarios, each covering a different part of it, so no score rests on a single example. Each skill is measured through five scenarios, with a bank of further scenarios behind them. |
The five dimensions of Leading with AI
Five skills passed all five gates. Each became a dimension: a distinct skill the assessment measures and scores on its own. Together they make up the rubric. Every scenario is written against it, and every score traces back to it.
Growing and applying own AI skills. The manager makes time to learn AI on real work, briefs it and gives it feedback as they would a team member, and checks its output before acting on it.
Enabling the team to use AI. They make AI use welcome through their own example, give clear permission and room to experiment, build their team’s skills, and spread what works.
Guiding people through the change. They explain why the team is adopting AI, respond honestly to concerns about roles, and tell different concerns apart, checking whether each holds before persuading or changing course.
Rethinking how the team works. They build AI into staffing and project plans, shift work to AI where it performs well, put freed time to good use, and reshape roles with the people in them.
Keeping quality and ownership. They judge AI’s gains on finished work, step back where it isn’t paying off, make sure a person owns every decision, and use only data and content they may share with AI.
Part two: how each scenario is built
Specifications set before writing
Before we write a scenario, we fix its specifications. They apply to the scenario and to every one of its options.
Audience. Managers whose teams already use, or are starting to use, AI, from team leads to executives, with some scenarios set in enablement, HR or learning and development roles.
Difficulty. The assessment is set at the intermediate level.
Plain language. US English and short sentences, so the scenario reads the same for every candidate, whatever their industry.
Tool-agnostic. No scenario depends on how a particular AI product works. Each one describes the AI in plain, vendor-neutral words.
Weighted options
Each scenario comes from a manager’s own work, with AI changing what the team does, and offers four options. Each option carries a weight from 4 to 1, and the weights follow the same logic in every scenario.
The best option leads well, in proportion to what’s at stake. It carries a weight of 4.
A workable option reaches an acceptable result, but at a real cost, like a lost week or a check aimed at the wrong risk. It carries a weight of 3.
A plausible option sounds reasonable, but rests on a misunderstanding of how AI changes the work. It carries a weight of 2.
The weakest option sounds sensible but leaves the problem in place, often because it is what a capable manager without experience of leading AI adoption would do. It carries a weight of 1.
In scenarios about guiding people through the change, the lower options are ordered by how well they lead people.
Every option comes with a written reason for its weight, and each less effective option reflects a different poor judgment. What earns the full weight also changes from one scenario to the next: sometimes it’s using more AI, sometimes stepping back, and sometimes acting on a concern a team member raises. So no single habit, like always picking the most cautious option, leads to a high score.
Gates for every scenario and option
Every scenario and every option goes through the same gates. Anything that fails is rewritten or cut.
Gate | The test |
|---|---|
Needs AI | Replace AI with any new software system, and the best action changes. |
The best move works | The best option resolves the problem the scenario sets. |
The weakest move falls short | The weakest option leaves the problem in place. |
No technical knowledge needed | The best action follows from the facts in the scenario and management judgment, not from knowing how AI tools work. |
Within a manager’s authority | The best option is something a manager in that role can actually do. |
Different poor judgments | Each less effective option reflects a different poor judgment, so no two can be ruled out for the same reason. |
No clue in the options | The four options are similar in length, open in a way that singles none out, and are each a single action. The best option is never the longest. |
How it’s scored
Each option’s weight counts toward its skill. Each of the five skills has its own score, and together they make an overall score. Scores are compared with norms and reported in four bands, from Poor to Exceptional, so the report shows where each manager is strong and where they need support. There’s no fixed pass mark: any cut-off is set for the role. Each scenario keeps a record of the difficulty it was set at and, once managers take it, the difficulty observed and the share choosing each option. We use those figures to keep revising the scenarios.
Frequently asked questions
How was Leading with AI built? From a job analysis of the work managers now do as AI changes their teams’ work, and published research on how people and teams take up new technology and AI. Subject-matter experts confirmed the skills, every skill passed five gates, and every scenario passed its own checks before use.
What does it measure? Five skills: growing and applying own AI skills, enabling the team to use AI, guiding people through the change, rethinking how the team works, and keeping quality and ownership. Each has its own score, and together they make an overall score.
How is it different from the AI Fluency Assessment? The AI Fluency Assessment measures how a person works with AI. Leading with AI measures how a manager leads a team through AI adoption. Any skill the general assessment already covers was left out.
Is it the same for candidates and current managers? Yes. The same assessment works for hiring, internal moves and leadership development.
