What Good Employers Commit to During an AI Transition

Most HR leaders want to handle AI change well. The pressure usually comes from somewhere else, such as a savings target or a competitor’s announcement. Meanwhile employee worry has climbed. Mercer’s Global Talent Trends 2026, a survey of nearly 12,000 executives, HR leaders, investors and employees worldwide, found that concern about losing a job to AI rose from 28% of employees in 2024 to 40% in 2026, and 62% of employees said leaders underestimate AI’s emotional impact, while only 19% of HR leaders build that impact into their implementation strategy (Mercer).

Writing commitments down before the plan is final makes the decent choice the default one, even when deadlines are tight. Below are eight commitments that hold up with employees and with regulators, along with examples of employers and unions that have put similar promises in writing.

Why commit to anything before the plan is final?

How people hear about change shapes what they do next. Gallup found that workers who come to expect their job will be eliminated report lower engagement and more burnout, and are 4.1 percentage points more likely to be actively job searching. Among frequent AI users, feeling cared for by the organization weakened the link between AI use and job-loss fear by 11.1 points (Gallup Workplace).

That gives a CHRO and a CFO a shared reason to act early. Commitments cost little to make before decisions are locked, and they protect the engagement of the people you plan to keep. With a record 27% of U.S. workers already worried that technology could make their jobs obsolete, silence tends to confirm the worry (Gallup).

1. How early should you tell employees about AI changes?

As early as you can say something true. Tell people which work you are testing AI on and when they will hear more, and be clear about what is decided and what is still open. A short, honest update every few weeks beats a polished announcement that arrives after the rumors.

Some employers have put this in writing. In its 2025 agreement with the Writers Guild of America East and West, CBS News committed to alert the union before rolling out new AI tools and to meet with the Guild twice a year to discuss generative AI (TheWrap). Microsoft’s 2023 partnership with the AFL-CIO commits to share in-depth information on AI trends with labor leaders and workers and to incorporate worker perspectives into AI development (Microsoft). For a step-by-step approach, see How to Tell Employees About AI.

2. Why should redeployment come before reduction?

Every person you redeploy keeps knowledge you would otherwise pay to rebuild. Executives already expect a lot of movement: in Mercer’s 2026 research, 65% expect 11% to 30% of their workforce to be redeployed or reskilled because of AI within two years. A practical commitment is that no one receives notice until their skills have been matched against open roles and the roles your plan says will grow, with internal candidates getting the first look. When Standard Chartered announced plans to cut more than 15% of back-office roles by 2030, it said affected employees would get retraining support or packages to help them move into other opportunities (Private Banker International).

Skipping this step can be expensive. In July 2025, Commonwealth Bank of Australia said it would replace 45 customer service roles with an AI voice bot. The union disputed the bank’s claim that call volumes had fallen and took the matter to the Fair Work Commission, and the bank reversed course, offering the workers their jobs back or a voluntary exit (American Banker, Cyber Daily).

Workers want a tech savvy bank, but they expect to be part of the change, not replaced by it.

3. How much time and money should retraining get?

Enough that people can finish it during working hours and land in a named role at the end. Employees value this highly: Mercer found that 63% would trade a hypothetical 10% pay raise for opportunities to build AI and digital skills, and 77% of investors said they are more likely to invest in companies committed to empowering employees through AI education and training. Microsoft’s AFL-CIO partnership included formal learning sessions for labor leaders and workers on how AI works, led by Microsoft’s AI experts (Microsoft).

Set the retraining budget in the same meeting as the AI budget, so the two move together. Measure it by the number of people placed in new roles, which tells you far more than hours of training completed.

4. What counts as fair notice and severance?

Start with the legal floor. Federal WARN requires 60 days’ written notice before covered plant closings and mass layoffs at employers with 100 or more employees (29 U.S.C. 2101, 2102). New York requires 90 days at employers with 50 or more full-time staff in the state. New Jersey requires employers with 100 or more employees to give 90 days plus severance of one week per full year of service, and four more weeks if notice falls short. The Fair Labor Standards Act itself contains no requirement for severance pay (U.S. Department of Labor).

Above the floor, recent packages show what large employers have offered.

5. Does outplacement make a difference?

It does for the people leaving, and it shapes what the people staying conclude about you. Challenger, Gray & Christmas, citing research in the Academy of Management Journal, notes that companies whose employees saw the layoff process as fair, consistent and well communicated had 48% less increase in voluntary turnover than companies whose employees did not (Challenger). Challenger sells outplacement services, so read that as a vendor citing research.

California now puts the question on the form. From 2026, WARN notices in the state must describe how the employer will support laid-off workers, including whether it will coordinate with the local workforce board (California EDD). A good commitment is outplacement for every affected person at every level, with enough paid time to use it.

6. What do managers need before an announcement?

Managers carry the message, and Gallup’s research suggests they shape how it lands. Employees who strongly agree that their organization cares about their wellbeing, or that they are respected, are about 6 to 7 points less likely to report displacement concern (Gallup Workplace). Commit that every manager with an affected team is briefed before any company-wide message, has a written talk track for each group, and knows where to send questions they cannot answer. Our piece on what employees fear about AI covers the questions managers hear most.

7. How do you avoid surprise announcements?

Set the order of communication and hold to it. Affected employees should hear from their own manager before an earnings call, a press release or a public filing tells them. Say plainly what role AI played. New York’s WARN filing has asked since March 2025 whether “technological innovation or automation” contributed to a layoff, and a public explanation that differs from the filing invites questions you do not want (Hunton Andrews Kurth).

8. Should you publish the plan and measure it?

At least internally, and on a schedule. NACD’s analysis of its 2026 board survey says that to strengthen oversight, boards need workforce reporting that goes “beyond headcount and turnover” to show capability gaps, reskilling needs and leadership pipeline risks (NACD). Yet only 19% of U.S. CHROs in a Conference Board survey said anticipated AI effects have been built into workforce and financial planning across the enterprise, according to the Conference Board. In New York, the legislature passed a bill in June 2026, A9581B, that would require publicly traded companies and employers with more than 50 people doing business in the state to report each year on AI’s effect on hiring and workforce reductions. As of late October 2026, the state’s bill page showed no action by the governor.

Pick a few measures and report them every quarter: the share of affected people redeployed, average time to placement, completion of retraining that leads to a new role, and how many eliminated roles were later rehired. Share them whether or not they flatter the plan.

Where should you start?

Most of these commitments depend on knowing, role by role, which tasks AI is taking on and where each person could move. JobRoute builds that view from public labor data, and we will prepare a free AI exposure sample for 10 of your roles in two business days so you can see it on your own job titles. Request the sample, or book a 15-minute call if you are shaping a transition plan now and want a second opinion.