McLean & Company published a guide on November 3 that tells HR leaders to define AI skills before building training around them. Gallup’s latest engagement data shows why one step has to come before even that. HR should write down what AI use looks like in each role, because only 46% of U.S. employees strongly agree they know what is expected of them at work.
What the two sources say
McLean & Company’s Define AI Skills Guide proposes four steps: establish the driving need and scope, build an AI skills library, review and prioritize the skills, and plan to embed them across the organization. The firm’s AI Skills Library has 19 foundational skills, nine technical and 10 durable. “Effective AI skills development can’t happen in the absence of an AI strategy,” said Lisa Highfield, principal director, HR Research & Advisory Services at McLean & Company, in the release.
Gallup’s third-quarter engagement report supplies the employee side of the same problem. The share who strongly agreed they knew what was expected of them fell from 49% to 46%, and engagement was 15 points higher in organizations that communicated a clear AI plan. I covered the full numbers in our report on the engagement drop.
My argument
A skills library defines what the organization wants employees to be able to do. Employees still have to learn what the organization wants them to do on Monday morning. Those are two documents, and most HR teams are building the first one. I think the second should be written first, role by role, in a page or less, and I think a skills library built without it will describe capabilities that nobody has been told to use.
Gallup’s own recommendation starts there. Its first priority for leaders is to clarify AI expectations by role, so that employees understand how AI applies to their work, when they should use it and how their performance will be evaluated as responsibilities change. The skills guide’s first step, clarifying why AI skills are needed and which roles matter most, points the same way. A role-level expectations page is the output of that step in a form an employee can read.
The strongest objection
The best case against me is that expectations written today will be out of date within a quarter. Tools change, vendors ship new agents and a page written in November may describe a workflow that no longer exists in February. McLean & Company makes the same point about skills, saying the library should be reviewed after major tool updates or changes to AI strategy.
That is a reason to keep the page short and dated. An expectations page with a review date is cheaper to revise than a skills library of 19 entries tailored to proficiency levels, and it gives managers something to say in the meantime. My earlier piece on skills profiles needing evidence of task performance argued for the same ordering in hiring, with evidence of the work coming before the profile.
What I would do this quarter
I would pick the three roles with the heaviest AI exposure and write one page for each. Each page would say which tasks AI is expected to handle, which tasks stay with the employee, who reviews AI output and how the work will be assessed. I would give the draft to the managers of hybrid and remote teams first, since Gallup found engagement fell five points among fully remote employees and two among hybrid employees in Q3, against about one point for on-site staff.
Then I would open the skills library. With the expectations in hand, the library has something to be checked against. A skill that appears in no role page is a skill nobody has been asked to use, and it can wait.
Each page can carry the same review date as the skills library, so the two documents are revised together after any major tool update. Managers then have one current answer when an employee asks how AI fits into their role.
The sequence costs a few weeks of manager time before any course is bought or built. By then HR will know which of the 19 skills, or whichever list it chooses, are tied to a task someone has been told to do.
