The Columbia Journalism Review recently published a detailed guide for newsrooms on how to develop AI policies. Reading it, one thing is clear: every challenge they describe also applies to internal communications teams.
Who is responsible when AI drafts an employee newsletter? What happens to trust when people realize their town hall summary was generated? How do you scale AI use without turning your comms into something that feels like it came from a content farm?
Internal comms leaders don’t have an AP Stylebook equivalent for AI yet. But the shape of good answers is starting to emerge.
What most teams get wrong at the start
Most teams that fail at AI policy start in the wrong place. They open with a list of approved tools, a set of dos and don’ts, and a ban on a few use cases. The technology changes and the list is stale before anyone reads it.
The organizations getting this right start differently. They ask: what do we actually believe about how we communicate with our employees? What standard of care do we owe them?
For an internal comms team, those values aren’t hard to find. Accuracy, trust, human connection, two-way dialogue. Any AI guideline worth following has to be grounded in those, not in a specific tool or use case.
Frame it as opportunity first
A policy that opens with eleven “do not” statements gets memorized for the compliance test and ignored in practice.
The Wall Street Journal’s editorial AI policy opens with “Advances in artificial intelligence create incredible opportunities for our newsroom.” The first restriction doesn’t appear for five paragraphs. That framing matters. It signals that leadership thinks AI is worth engaging with, not just worth policing.
Internal comms teams building their own guidelines should take the same approach. Name the use cases AI makes genuinely easier: summarizing long leadership recordings, translating content for global teams, drafting first-pass newsletters for editor review. Put those upfront. The guardrails are more credible when people understand what they’re protecting.
A policy document without training does nothing
FT Strategies research found that 61% of media organizations cite skills gaps as the primary barrier to AI adoption. Cultural resistance came second at 52%. Unclear use cases third at 45%. The same distribution exists in most enterprise comms functions.
Reuters runs mandatory AI training for all staff annually, plus monthly town halls and regular hands-on demos. Zetland built an internal transcription tool and let the positive word-of-mouth do the adoption work for them.
The document is the least important part of the process. The culture around AI use matters far more. People need to know how to act when a specific situation comes up, not just that a policy exists.
Transparency with employees is non-negotiable
Newsrooms debate whether to label AI-assisted content. The internal comms answer is more straightforward: employees notice when messaging feels generic. They may not know why, but they feel it. Once trust in internal comms erodes, rebuilding it is slow.
Two defensible positions: disclose significant AI use in the communication itself, or treat the human author as fully responsible for everything that bears their name (AI assists, an editor publishes, the editor owns it). Each works. What doesn’t work is the default most teams end up at — informal AI use, no clear standard, no training, and the assumption that no one will ask.
Build it with people, not for them
Sahan Journal drafted their AI policy through a committee, one-on-one sessions with department heads, an all-staff meeting, and open office hours for feedback. The result was a policy people knew existed and understood.
Internal comms teams have the tools for this already. Pulse surveys, focus groups, town halls. Use the same infrastructure you’d use to understand what employees need to involve them in shaping rules that affect how their company communicates with them.
Build in a review trigger, not a review date
AI moves fast. A policy drafted in 2024 is already missing several relevant use cases. Zetland’s CEO framed it well: “Don’t overthink, and don’t be too concrete. Rules, model recommendations, and dos and don’ts all quickly become outdated. Principles should last longer.”
The right structure: stable principles at the top, practical guidance in the middle, specific tool and use-case notes that update as needed. The review trigger is a condition, not a calendar date. When a new capability appears that your current policy doesn’t address, that’s when you update.
Internal communications is one of the few functions in an organization where trust is the product. Newsletters, town halls, manager updates, crisis communications — all of it depends on employees believing what they’re reading is honest and worth their attention.
AI can help a team of two communicate across a company of two thousand. It can reduce the manual work that keeps good content from getting made. But that only holds if the team makes deliberate decisions about where AI belongs in the process. The newsrooms that got this right weren’t faster or slower than others. They were more explicit about what they were protecting.
