Executive Summary
Leadership decisions that shape participation in AI transformation.
Leaders should examine staffing plans, workload expectations and recognition before treating limited AI use as an employee capability problem. Many concerns in this paper arise from decisions executives control: what happens to staffing, who benefits from improvements, how performance is judged and what employee data is used for. Resolving these concerns requires leaders to make and explain those decisions.
These concerns affect what the business learns from its investment. Employees who expect higher workloads or fewer jobs have reason to keep useful discoveries to themselves. Managers accountable for unchanged targets have little room to support experimentation. Executives who approve investments without understanding the work are poorly placed to judge the support people need or the results to expect.
Eight groups of concern
The paper sets out eight groups of reasons employees may hesitate to use AI at work and who holds the main authority to respond to each. Most sit with executive decisions on employment terms, policy and investment.
| Group | Concerns covered | Who can change the conditions |
|---|---|---|
| Economic fear | Job loss; who keeps the gain; training their replacement | Executive with HR and finance: staffing assumptions, use of gains, recognition and redeployment. |
| Identity and skill | Pride in the craft; losing skills; loss of human contact | Executive and manager: valued expertise, professional standards, human contact and development. Employees contribute professional judgment. |
| Social risk | Looking inexperienced while learning; stigma; credit | Executive and manager: credit, evaluation and responses to questions. Employees support colleagues and follow agreed disclosure rules. |
| Trust in the output | Accuracy; accountability | Business owner with technical and risk teams: quality, review and accountability. Employees check outputs and escalate failures. |
| Practical friction | Slower for the actual task; poor fit with existing tools; no time or support to learn | Manager and system owner within an executive budget: workflow fit, access, integration and practice time. Employees test real tasks. |
| Rules and risk | Unclear policy; confidentiality; surveillance | Executive with security, privacy, legal and HR: approved data use and monitoring. Managers apply the rules; employees follow them. |
| Values | Ethical objections; distrust of vendors | Accountable executive with relevant specialists: acceptable uses, vendor requirements and human involvement. Employees raise objections. |
| Signals from the organization | Leaders do not engage with the tools; incentives go the other way; mixed messages | Executive leadership team: consistent staffing messages, incentives, budgets and conduct. Managers raise unresolved conflicts. |
What leaders and managers need to address
State the staffing plan and how gains will be used
Before asking a team to redesign its work, the executive sponsor and HR leader should state whether the business case assumes reduced headcount, slower hiring, redeployment or additional capacity, and give the next decision date where the plan is unresolved.
Change the measures when the work changes
If AI removes routine customer inquiries, the service leader should approve measures suited to the remaining complex cases before rollout, including which handling-time target changes and how much coaching time managers receive.
Publish the rules for use and employee data
State approved tasks and information, required review and escalation routes, and what prompts and activity data are recorded, who can inspect them and whether they enter performance decisions.
Protect learning time and define competent performance
Fund practice within working hours, decide which output expectations are adjusted during it, and specify how AI-assisted work earns credit. Managers and executives take part in the same practical learning.
What the research supports
Across experiments involving more than 4,400 participants, researchers at Duke found that people who use AI can be judged less favorably on competence and motivation, and that people anticipate that judgment. A study of 5,172 customer-support agents found a 15% average increase in issues resolved per hour with AI assistance, with the largest benefits for less experienced staff. The paper sets out what each study does and does not establish.
Topics
The full paper covers the eight groups of concern, what the research supports, the four decisions leaders and managers need to make and an example of what stays unknown when useful practices go unshared. Free, no gate.