Most companies are experiencing some level of turmoil over their adoption of Artificial Intelligence. They should be able to lean on their managers to interpret these issues at the ground level. Research, however, finds that training has left managers out. They may be getting some of the same training all employees are getting, but nothing to help them guide their teams. That will become more and more problematic as challenges continue to mount—like the two issues that have emerged recently: companies shifting the models they’re using, and employees stepping out of their areas of expertise because AI can give them information they would normally turn to internal subject matter experts for.

Links from this episode:

The next monthly, long-form episode of FIR is tentatively scheduled to drop on Monday, August 24.

We host a Communicators Zoom Chat most Thursdays at 1 p.m. ET. To obtain the credentials needed to participate, contact Shel or Neville directly, request them in our Facebook group, or email fircomments@gmail.com.

Special thanks to Jay Moonah for the opening and closing music.

You can find the stories from which Shel’s FIR content is selected at Shel’s Link Blog. You can catch up with both co-hosts on Neville’s blog and Shel’s blog.

Disclaimer: The opinions expressed in this podcast are Shel’s and Neville’s and do not reflect the views of their employers and/or clients.

Raw Transcript:

Shel Holtz:

When companies change something they’ve been extolling for the last six months, managers are expected to support that change. When employees step outside their lane, managers are supposed to manage that. But can managers do these things when they haven’t been prepared?

That’s the case with some of the fallout from the introduction of artificial intelligence in organizations. The details are coming up in this short midweek episode of For Immediate Release.

Hi, everybody, and welcome to episode 524 of For Immediate Release. I’m Shel Holtz, and you’ve got just me again this week.

I have three artificial intelligence stories to share with you. Organizational communicators are going to have to live at the intersection of all three of these issues, whether we’re ready or not, so let’s tackle them as parts of a bigger whole.

I’ll dive into these stories right after this.

Let’s start with the whiplash.

For the past year or so, companies have been in what some have come to call the “token-maxing era”: Throw AI at everything, reward employees for burning through as many tokens as possible, and worry about the ROI later.

Nvidia’s Jensen Huang was quoted as saying that if your half-million-dollar engineer isn’t burning $250,000 in tokens, something’s amiss. Meta reportedly ran an internal competition rewarding token usage.

That was the mood last spring.

Summer’s mood is different.

The Associated Press—and, by the way, this same reporting ran in The Washington Post, The Philadelphia Inquirer, and elsewhere—describes companies now hitting a wall. Costs went up, productivity didn’t keep pace, and the bill is coming due.

We covered this in episode 517 back in early June, but just as a reminder: Uber blew through its entire annual AI budget in four months, according to its CTO, and has since rolled out spending tiers starting at $1,500 a month per employee.

Some companies aren’t just tightening budgets; they’re switching platforms entirely. One AI startup called Lindy moved 100% of its traffic off Claude and onto DeepSeek, the cheaper Chinese alternative. Its CEO told CNBC that the cost curve “crashed to the ground.”

Multiply that kind of switch across every company currently having the same conversation, and you’ve got a specific communication problem: How do you explain to a workforce that, after spending months enthusiastically evangelizing one tool, you’re now moving employees to a different one—possibly overnight—for reasons that are mostly financial and don’t actually involve the employees who have been building things on the original tool?

That’s problem number one.

Problem number two is going to make problem number one look simple, because it’s not just about which tool people use. It’s about what people are allowed to do with it.

Axios got an exclusive look at new OpenAI research drawn from more than 800,000 work-related ChatGPT messages from business users. The finding is that workers are routinely doing other people’s jobs.

Roughly 44% of occupation-specific requests involve tasks typically associated with a different profession. After stripping out generic activities such as drafting emails, customer service employees, designers, and HR professionals are the biggest boundary crossers. Something like three-quarters of their profession-specific prompts touch work that belongs to somebody else’s job title.

People are using ChatGPT to draft marketing materials, troubleshoot software, run financial calculations, and interpret regulations—jobs that used to require reaching out to a specialist for help.

OpenAI’s chief economist told Axios that the boundaries between jobs are already becoming more flexible because of this.

Now, if you’re in communications, you can probably already hear the governance questions stacking up.

Who’s accountable when a well-meaning employee uses AI to draft something that touches legal, financial, or regulatory territory in which they have no training?

What happens to your carefully built subject-matter-expert review process when everyone feels like a generalist?

And when something goes wrong because of bad advice, off-brand sentiment, or a compliance miss, who owns that failure? Is it the employee, the tool, or the manager for not seeing it coming?

That brings us to the third piece, and honestly, it’s the one that worries me most.

HR Dive reported on new research from Indeed and YouGov. The headline number is that 43% of managers say they feel poorly equipped—or not equipped at all—to lead a workforce that’s fluent in AI.

More than half of workers say they’re not getting the AI training they need. Employer expectations for what workers should be doing with AI are running two or three times ahead of what workers actually feel comfortable doing.

Indeed pointed out that companies have spent two years building AI-ready workforces. The next challenge—and arguably the harder one—is building AI-ready leaders.

Now, put these three stories together and you get a pretty clear picture.

Companies are going to keep switching models and vendors as the economics shift. Employees are going to keep wandering across job boundaries because the tools make that easy and, honestly, tempting.

The people standing in the middle of both trends—the frontline managers who must explain the switch, catch the boundary problems, and answer the “Wait, am I even allowed to do this?” questions in real time—are the group companies have prepared the least.

That is a communication problem, and it’s ours to fix.

If managers don’t understand why the company moved off the model they spent six months championing, they can’t credibly explain it to their teams. They’ll either go silent, which breeds suspicion, or they’ll improvise, which is worse.

If managers don’t have clear guardrails explaining what kinds of AI-assisted work outside someone’s lane are acceptable and what kinds require a specialist’s involvement, they’ll either rubber-stamp everything or block everything. Neither option serves the business.

And if leadership training lags this far behind workforce adoption, managers become the bottleneck at exactly the moment the company needs them to be the translators and interpreters.

Here are three things I’d be pushing for internally right now:

First, build the change narrative for platform and model switches before you need it. Develop a plain...