Field observations from conversations with enterprise leaders.
In recent months, we’ve talked with leaders at some of the world’s largest companies, from the factory floor to the store shelf to the server room. The conversations were supposed to be about communication, but nearly all of them turned into discussions about AI.
AI has made content cheap to produce, but it hasn’t made it any cheaper to understand, judge, or act on. While companies moved fast to hand people AI tools, they haven’t moved nearly as fast to agree on how to use them, what good looks like, and who’s responsible when the output falls short.
We heard leaders describe this tension from every side:
- The people asked to use AI every day: “We’re all being asked to use it, but nobody quite knows how.”
- The people on the receiving end, who sit through long decks that all look alike.
- The people trying to make adoption stick, who need proof that AI is changing the work, not just more logins.
Their frustrations look different, but each group needs the same thing: a common standard for turning information into a decision.
Where companies are getting stuck
Here’s what we heard most often, and the pattern underneath each problem:
Wide adoption, shallow impact. Most companies have finished the horizontal phase of AI: access, some training, and a push to use it. The vertical phase of embedding AI into business process hasn’t started in earnest. One company wants advanced AI use on every scorecard, yet its own leaders admit most people are “using it for an email.” Another wants tools “designed into process updates and not bolted on,” but sees progress only “in pockets.” Horizontal adoption produces activity, while vertical adoption produces value. The worst case, one leader said, is “Excel macro on steroids… 7,000 versions of the same thing.”
The pattern: The Horizontal Plateau. Access is mastered, but the work hasn’t changed.
Shared tools, no shared rules. Without agreed-upon guidelines, each person decides what needs review, how facts get checked, and what’s safe to share. Most companies are still early. Some haven’t set AI standards for privacy or outside vendors. Others are boiling guardrails down to “five things you cannot do.” Governance, where it exists, covers data and compliance but rarely communication, which is where most AI output lands. People either avoid AI or over-trust it and stop applying judgment. Either way, inconsistency grows.
The pattern: The Guardrail Vacuum. Where no rules exist, habits become the rules.
Polished slides, outsourced thinking. “People are now not even thinking. They’re just outsourcing their brain,” one leader told us. The old problem was slides stuffed with pasted text. The new one is clean slides with no thinking behind them. The problem just changed clothes. Presentations become “a cleaner data dump,” and one CEO’s verdict was “This could have been an email.”
The pattern: The Information-Action Gap. AI makes it easy to report the news, but not to drive a decision.
More content, less trust. Documents are getting longer, and audiences are tuning out: “The minute I see a slide, I know it’s coming from AI, and I don’t want to pay too much attention to that.” Accuracy suffers too. Some leaders are seeing AI-generated slides that misrepresent their own products.
The pattern: The Credibility Discount. Polish used to signal effort. Now it can signal the opposite.
What’s at stake? The real risk isn’t an occasional wrong answer. It’s thousands of people quietly inventing thousands of standards. Decisions slow, leaders rewrite more, confidential information leaks, and the brand sits in the hands of whoever hit “go” last.
What’s working
The companies making real progress with AI aren’t the ones with the most tools. They share these habits.
They keep the rules simple and human. A short list beats a 40-page policy. One company uses a one-way door, two-way door test: decisions that can’t be undone get more scrutiny of the AI output behind them.
They make people think before they prompt. One internal AI agent asks five or six questions before it gives an answer, “prompting you to think as opposed to just prompting it.”
They define what good looks like, once. AI output changes with every person, prompt, and tool. Consistency needs a shared standard: know the audience, name the conflict, state the big idea, recommend the action.
They keep the human in the driver’s seat. Treat AI like a capable intern that handles speed. The person owns the thinking: what matters, what the data means, what should happen next. As one leader put it, “The ultimate human accountability doesn’t change.” But accountability only works when everyone shares the same definition of good.
They give the framework to everyone, not a select few. Communication is a team sport. Whoever pulls the data, shapes the story, or presents it touches the narrative. If only one person knows the framework, the story breaks at every hand-off.
Why storytelling is the guardrail
Most AI output ends up as communication. That makes a common storytelling framework, given to the whole organization, the answer to the guardrail gap. It gives everyone the same definition of good, the same way to review AI output, and the same language for pushing back when something looks polished but lacks substance. A shared framework doesn’t slow AI down; rather, it gives AI something to aim at.
Companies have already rolled out AI at scale. Now they need to roll out the guardrails at scale, too. How? By delivering a short, foundational storytelling course everyone takes in the flow of work, reinforced by leaders. That’s how a single training event becomes a lasting guardrail across the company.
The first wave of AI was about access, and the next is about judgment. Tools can be standardized with a purchase order. But judgment can’t be bought. It has to be taught and reinforced, or it splinters into thousands of private versions.
AI didn’t lower the bar for communication. It hid it. Companies that make it visible again, for everyone, will get the return on tools they’ve already bought.
Key takeaways
- Access isn’t adoption. The value is in embedding AI into real workflows.
- Without shared guidelines, every employee writes their own.
- Polished output is no longer evidence of thinking.
- AI handles speed. People own judgment.
- A common storytelling framework, given to everyone, is the most practical AI guardrail.