Why your strategy might die in the AI-first intranet

Last week, I was interviewed by my laptop for three hours.   

I set up Claude Cowork and answered 100 questions. How I write. How I decide. What I can’t stand. The result was two files: an “About me” and a “Writing style” md-file, now stored permanently as context for everything I run in Claude. Three hours, just so the AI sounds more like me, not a generic model. Because if you don’t give the AI your context, it fills the gap with a generic version of you. That’s where narrative drift begins: what happens when AI compresses your strategy into a generic version of itself before a single human reads it. 

What’s a personal profile for me is your strategy at the company level. For decades, strategy communication was human-first. People read it, made sense of it, passed it on. In 2026 it’s different. AI reads your content before your employees do, summarises it and serves it back through a chatbot or a search box. And increasingly, AI helps create the strategy in the first place.  

On stage at a recent communications conference, I ran an example. A company announces a radical shift to a subscription model. Press release, intranet article, manager briefing, all drafted with AI. Then I asked the same AI to summarise them. Slightly smoothed in the writing, fully flattened in the reading. What’s left, three times over, is the same wishy washy “we want to become more agile and customer-focused.” The actual strategy shift was almost completely gone. 

AI compresses by statistical probability, not by strategic priority. Even before AI, strategy barely reached “the bottom”. According to Harvard Business Review, only half of middle managers can name even one of their company’s top five priorities. Among frontline staff, it’s only 16 percent. Telephone game at scale. And now AI stacks even two more layers of compression on top.

A Staffbase-YouGov study of UK employees found that when leadership communication about vision and strategy is very clear, 92% report being very or somewhat happy in their jobs. When it’s very unclear, only 25% feel that way. That gap, between a strategy that survives compression and one that doesn’t, is the cost of narrative drift. 

Since Claude Shannon in 1948 we’ve known that compression without error correction ends in noise. When models summarise, they pick the most probable and therefore most average path.

The Financial Times showed in March that where social media amplifies fringe views, AI nudges users toward the center. AI finds compromises, but your sharp original turns into a generic version of your company. AI slop. 

Don’t get me wrong. AI is a huge opportunity on both ends, creating and consuming content. Above all for communicating more personalised at scale. But just because we can produce more, faster, doesn’t mean we should without solid foundations like governed content, the right context, and a structure free of contradictions. Under time pressure, content slips into mediocre quality.  

Your intranet is now the source the AI pulls its answers from. It constantly recombines your content, and the structure decides the quality of the answer. Almost right is wrong at work. 

So what helps?  

• Tidy up and restructure. Only structured content without contradictions survives compression. 

• Give the AI the context it doesn’t have on its own. Exactly what I did last week. 

• Write your strategy content to be “durable, sturdy enough that it still carries your core message after the machine summarises it. 

Good communication in the AI age also means making sure your strategy survives the machine that summarises it before a human ever sees it.