Where Should AI End and Leadership Begin? Three Lines Communication Experts Won’t Cross

28 Apr Where Should AI End and Leadership Begin? Three Lines Communication Experts Won’t Cross

This is the first in a series on leadership communication in a changing landscape, drawn from thousands of client conversations. More than most, Speakeasy appreciates that this territory is nuanced. What follows is what our experts are seeing.

Leaders are using AI to communicate. That’s not the debate. The harder question—the one most organizations haven’t answered yet—is what you can hand to AI and what you still have to own yourself.

We put that question to three members of Speakeasy’s faculty: Daniel Frysh, Ed Domansky and Sandra Ashe. Between them, they’ve coached senior leaders through some of the most consequential communication moments of their careers.

Coming at it from three different angles, they drew three different lines. Here’s where each of them lands.

Daniel Frysh: Never Let AI Write First

Daniel put it the most directly: judgment comes first, always. AI is a tool—a good one, a necessary one—but it cannot “superimpose itself over our own intelligence.” The moment it does, he argues, two things go wrong at once.

First, it sounds fake. Audiences can feel the uncanniness of the message, even when they can’t name it. In Daniel’s words, it comes across as “wildly inauthentic.”

Second, AI isn’t a neutral mirror. It has a documented habit of inventing plausible-sounding ideas it thinks the user wants to hear—a sycophancy problem Daniel called the “hallucinatory moment.” When a leader hands over the creative step, they’re not outsourcing to a reliable assistant; they’re outsourcing to something that will cheerfully make things up to please them.

Daniel’s rule is clean: creation is the communication that should never be delegated to AI. Once you’ve done the thinking and drafted the substance yourself, AI earns a seat at the table. It can polish. It can play devil’s advocate. It can stress-test whether you’ve covered the full scope of an argument. But it cannot originate.

He added a point the others didn’t: a leader’s responsibility on an AI-enabled team is to know their people well enough to know which tools fit which individuals. Not every platform serves every person the same way. Matching the tool to the human is itself a leadership act.

Ed Domansky: AI Gets You Started. You Finish.

Ed, drawing on his corporate communications background, came at the same question from the quality-control side.

Leaders today, he said, tend to use AI “almost as a crutch” in place of their own professional judgment. He was careful not to demonize the tool.

AI is a great starting point for breaking writer’s block, for getting a leader unstuck, for generating the first thoughts on the page.

The failure isn’t in the use. It’s in what comes next.

After the machine gives you something, the real work must start.  Ed laid out the checklist:

• Is it aligned with your audience? Does the message properly target the reality of the people you’re trying to reach?
• Is the tone right? Does it sound like you and what the moment calls for? Will it be received the way you want and need it to be received by the target audience?
• Is it accurate? This, for Ed, is the hugest thing.

On accuracy, he was specific about the risk.

AI pulls from so many sources that its output is “a mix and match”—plausible-looking synthesis that may not survive a closer look. Which means the expert (you) must go back through the output and verify that it’s correct and will land the way you need it to.

Ed’s stance is pragmatic: You can use AI to get started, but you cannot rely on AI to do all the work.  The edit, the fact-check, and the needs of the audience are critical elements that  leaders must understand and own.

Sandra Ashe: “Anything Goes” Is Not a Policy

Sandra stepped back and drew the line at the organizational level.

Where Daniel and Ed focused on the individual leader’s discipline, Sandra pointed to the systems around the leader. Most companies, she observed, lack meaningful governance around AI use. At this point in our adoption of AI, she said, “anything goes right now,” and the abuses are already visible.

She’s not exaggerating. Recent research from the Thomson Reuters Foundation found that only 38% of US companies have even published an AI policy, meaning most leaders are operating in exactly the “free for all” Sandra describes.

That matters because individual good judgment only carries an organization so far. Without shared rules, one leader’s careful stewardship lives next door to another leader’s fully AI-generated client email, and the business’s voice fractures.

Responsible use, in Sandra’s framing, looks like three things operating together:

1. Strong governance and safety requirements around how AI is used.
2. Monitoring how employees are using it.
3. A clear sense of where AI genuinely adds value—which, for her, is checking information and going deeper into the assumptions we make about our listeners.

Like Daniel and Ed, Sandra named a place where AI cannot go. For her, that place is personal communication. Employees and clients need to feel something unique from a leader, something that sounds like it came from the leader’s own head and heart. If it sounds as though it was lifted off an AI platform, she said, it never comes across as trustworthy.

What They Refuse to Give Up

Three faculty. Three places to draw the line. But one principle sits underneath all three answers:

Accountability can’t be delegated.

AI doesn’t take the meeting. AI doesn’t face the employee when the news is bad. AI doesn’t sit across from the client when trust has to be rebuilt.

The leader does.

And when it goes sideways—when the email lands wrong, when the all-hands message rings hollow— no one is going to care which model you used. That’s the piece AI cannot absorb, no matter how polished the output gets.

A leader can use AI to prepare for the moment. The moment itself still belongs to them.

Where’s the line in your own work—the place where, no matter how good AI gets, you won’t hand it off?

Daniel Frysh, Ed Domansky, and Sandra Ashe are members of the Speakeasy faculty. Speakeasy has coached senior leaders through high-stakes communication moments for more than 50 years.