PR Teams Split Tasks With AI, but Message Drifts at Every Handoff

By James Calloway

A single sentence can leave the building and come back meaning something else. Chris Gee, founder of Chris Gee Consulting and an advisor for Ragan's Center for AI Strategy, says AI accelerates that drift — and most communications teams haven't built a way to catch it.

The problem is familiar. A team spends weeks on the mission line, the vision statement, the phrasing legal approved and leadership likes. Then a colleague edits it for a partner newsletter, someone trims it for a board memo, and a third person rewrites it for a policymaker's staff.

"That's where AI helps most," Gee writes. "It's also where meaning starts to slowly drift."

Why AI handoffs are a telephone game

Gee reaches for an old analogy: the telephone game. Each whisper changed the message slightly, and nobody noticed until the last kid spoke. "AI makes each handoff faster," he writes. "It doesn't guarantee that the message survives the journey between hands."

He watched the dynamic play out in a recent workshop with a communications team. The team sorted five recurring tasks into three piles: do it myself, do it with AI's help, or hand it off entirely.

The votes mostly lined up. The reasons behind them did not — and those reasons clustered around four questions.

What questions should govern AI use?

Gee's workshop surfaced four tests for deciding when AI touches approved language:

• If it goes wrong, can it be fixed? A newsletter mistake may go unnoticed and gets corrected next issue. A leader's mistake in front of an audience can damage the company's reputation.
• Whose voice is it? Short news blurbs can be AI-assisted, but anything that influences someone's point of view must stay in human hands.
• Is it original or repurposed? Reworking approved material carries far less risk than writing from a blank page.
• Does doing it by hand build a skill your team needs? Writing skill and critical thinking came up first. One participant said she'd rather read a full report than ask AI to summarize it, because that's how she learns what makes her audience tick.

The takeaway: reversibility, voice, provenance and skill-building decide where the AI line sits.

Where does the split actually land?

Most tasks broke in half, Gee reports. AI can pull the analytics numbers, but a person must explain what they mean. AI could find every time a CEO has talked about an issue by analyzing hours of transcripts, but a person must decide what really matters from that data.

"Use AI at any point in the work but keep the decisions that go out under your name," Gee writes. "Think copilot, not autopilot."

In plain terms: machine does the retrieval, human keeps the judgment.

The one-week exercise

Gee prescribes a concrete test. This week, pick one core message your team owns. Ask your AI tool to adapt it for three audiences — a partner, the board, and a policymaker. Then put all three versions next to the original.

Mark every line where the meaning moved away from its original intent. Those marks become the review checklist. Share it with the team and run every altered piece against it before publication.

"Your messaging guide may be built to travel, but you have to know where to check it at each stop," Gee writes.

For teams under pressure to ship AI-assisted copy faster, the exercise turns drift from an invisible risk into a named, checkable list — before a regulator, a board member or a journalist finds it first.