McKinsey asked employees which company moves would increase their daily use of generative AI tools. Formal training from the organisation drew the largest share. For finance, HR and operations, that result is a brief for the first lesson, and for the check a manager can run afterwards.
The finding is in McKinsey's Superagency in the Workplace report. The underlying survey, run in October and November 2024, covered 3,613 employees and 238 C-level executives. Employees included people in finance, along with business development, marketing, product management, sales and technology. Among US employees, 48 percent said formal generative AI training from their organisation would increase daily use. Forty-five percent said the same of seamless integration into the workflows they already run. A requirement to use the tools for a certificate sat at 30 percent. Twenty-two percent said the support they currently receive for building this capability is minimal or none.
What to teach before anything else
The same report says 46 percent of leaders see skill gaps as a significant barrier to adoption. The remedy it points to is training that matches the role. Technical staff may need a technical bootcamp. Functional teams need practice at prompting, which McKinsey defines as designing and refining the input so the model produces the output you actually wanted. That is the first lesson for finance, HR and operations, because a tour of buttons or a session built around a made-up email gives a payroll analyst or a roster clerk nothing they can repeat on Thursday.
Pick one task each function already owes. In finance, a variance comment, or a clause taken from a contract the team is allowed to open. In HR, a first answer to a policy question that still has to match the handbook. In operations, a handover note or a supplier chase that already has a house format. The person writes what they are trying to produce, the constraints that bind the answer, and the context the tool does not have. A constraint might be a figure they must take from the source file, a tone the function already uses, or a rule they are not allowed to relax. They then read the result against that source and mark what they would send and what they would correct. What they are practising is a request that carries the job, together with a human check on the parts that are easy to get wrong, such as a number, a name, or a policy line.
How a manager can tell
McKinsey's employees ranked training and workflow integration above a certificate, a manager's instruction, or a KPI tied to use. Two weeks after the session, look at the task you named, in the function that owns it. Is the tool being used on that task, inside the way the work already moves? Is the check happening before anything goes to a colleague, a candidate, a supplier, or a customer? Can the person show the request, including the constraint, and the edit they made when the draft was fluent and a fact was wrong? An attendance sheet tells you who sat in the room, which leaves the question of the task unanswered.
Collective Campus builds high-performance teams for the AI world. The AI training workshop runs this practice on tasks your finance, HR and operations people already own, with the person who reviews the work in the room so the standard is visible. The version that goes deeper on the finance close is in what AI training for finance teams should change in the close. What a single corporate day should leave a team able to do is in what a one-day corporate AI training session should leave the team able to do.
Bring one live task from each function, the name of the person who checks it, and the tool people are allowed to open. We will shape the day around those three tasks. The following fortnight is the measure McKinsey's own respondents pointed at: use on the real workflow, and a check a colleague can see.
