When Your Stakeholders Can't Keep Up

The Federated Reality

Hearst Corporation spans 20,000 employees across hundreds of businesses—newspapers, magazines, television stations, data services, and digital platforms. When CEO Steve Swartz announced enterprise-wide AI upskilling in January 2024, the federated structure immediately complicated execution. BUs run their own P&Ls, and local leaders hold real power, which contracted the target audience from 20,000 to 10,000.

Early resistance varied by function. Newsrooms worried about journalistic integrity. Magazines insisted creative work remained human territory. The surveillance concern cut across groups—employees questioning whether their AI usage would be monitored.

We have trust, but we have a federated system. People trust local leaders, so we built communication through business leaders and local HR, not just from the center.

This is the kind of power mapping that effective ​change definition​ requires—understanding where authority actually resides, what other changes compete for attention, and where emotional resistance will concentrate.

Reframing AI as Skill, Not Threat

The initial "transformation" messaging failed. The team regrouped and shifted their approach.

"We started the introduction way more smoothly, saying, we understand, things are changing, things are moving fast, this is a new skill... it might be scary, but this is one skill in your skill box," Maris said. The team encouraged experimentation without immediate pressure.

Even if you don't find a use case immediately... Find something where you can just try it out, just get comfortable.

The reframe worked because it removed existential threat. Instead of selling organizational transformation, they positioned AI literacy as career development—something that increases individual agency rather than replacing human expertise.

The team also realized there were overlapping changes affecting different groups. This included external events like the Hollywood writers’ strike, or internal initiatives like reorganizations or system upgrades. Most change programs ignore these concurrent disruptions. Hearst's team explicitly acknowledged them and adjusted messaging accordingly.

Context-setting matters because leaders and employees experience change at different speeds. Execs announcing ​an initiative​ have already processed their own concerns. Employees are just beginning that journey.

The Speed Problem

The Chief AI and Product Officer came from startup environments and set what felt like an impossible deadline: 7,000 people through customized training by July 4th, starting in early May. Maris said:

His deadlines felt unrealistic to people from L&D, but he was not wrong. Aiming for perfect means months of polishing, and then coming out with generative AI after months, it's already outdated.

The demands included 107 live sessions across 16 functional groups, customized content for each business unit, and external facilitators. Three functional groups and 40 session dates were added the Thursday before Monday launch.

The speed created unusual urgency for an L&D function. More significantly, the compressed timeline exposed a structural gap in the ​change program​: the L&D team found themselves simultaneously building content and selling resistant business unit leaders on participation—operating as both change agents and de facto sponsors.

When Execution Outpaces Stakeholders

At Hearst, the compressed timeline forced the L&D team to operate as both change agents and de facto sponsors. They built local communication, created function-specific use cases, and demonstrated relevance before resistance could solidify.

Most importantly, they launched an AI Champions network from the start...though it looked less polished than most formal programs. "At the beginning we didn't have a regular process. It was a grassroots movement with uneven commitments," Maris clarified. "We didn't always have time to support the AI Champions, except providing some additional resources and access to our Chief AI Strategist."

The champion approach succeeded because it gave local advocates permission to act. In federated organizations, that permission matters more than perfect process. These champions provided what the central team couldn't: credibility within their own business units.

The strategy worked, according to several metrics:

• 7,000 employees completed training (the 70% goal)
• Active HearstGPT usage increased 175%
• Business conversations with the Chief AI Officer grew 200%

"[The Chief AI Officer] was restless at the beginning because he didn't have many business conversations," Maris explained. "In the middle of the program, he was like, I'm too busy."

The metrics indicate success. But they don't capture the strain on the execution team or address the underlying structural gap: business unit leaders who should have been actively ​sponsoring​ the change bought in only days before launch. The L&D team compensated brilliantly. That compensation doesn't make the model optimal.