Most organizations I work with already have AI in production. Dozens of pilots, a copilot rollout, an internal chatbot, maybe an agent or two. What they do not have is a clear answer when I ask the CEO and a frontline employee, separately, what the AI transformation is actually for. Two different sentences come back. Often two different worlds.
That gap is not a tooling problem. The tools are fine. The narrative is missing.
Three studies, one number
Three independent reports last year reached the same conclusion from completely different methodologies.
- BCG surveyed 1,250 senior executives across nine industries and found that 5% of companies are "future-built," generating 1.7x the revenue growth and 1.6x the EBIT margin of their peers from AI investment.
- McKinsey's State of AI 2025 found that ~6% of organisations are high performers attributing 5% or more of EBIT to AI, while 88% have AI in production somewhere and 94% report not seeing significant value.
- MIT NANDA's GenAI Divide study of 300 deployments concluded that 95% of GenAI pilots fail to scale, with cost overruns averaging 380% at production versus pilot projections.
Three different studies. The same number for the winners.
The standard explanation for that gap is technology, with model quality and integration fragility usually leading the list. The data says something different. RAND's 2025 review of more than a thousand AI projects found 63% of implementation challenges stem from human factors, not technical limitations. Projects with dedicated change management succeeded at 2.9x the rate of those without.
The 94-versus-6 gap between AI in production and AI capturing value is, before anything else, a narrative gap. This is the same gap I described last issue as the dual mandate HR has to lead, and the same one Tobi Lütke's Shopify memo and Satya Nadella's "bicycles for the mind" framing are trying to close in their own voices. Different shapes, same fight.
Andy Raskin's strategic narrative framework
Andy Raskin has spent a decade helping CEOs at companies including Salesforce, Square, Uber, Gong, and Dropbox build what he calls strategic narrative: a single story the CEO uses to align marketing, sales, product, fundraising, and recruiting around one direction. His framework has five pieces.
1. Name the shift. Old game to new game. Salesforce told customers "software is over, the cloud is coming." Zuora told them "transactions are over, subscriptions are coming." Gong told them "opinions are over, reality is coming." Each is one sentence, deliberately compact. Raskin's discipline: concise beats complete.
2. Name the stakes. Show that the winners are already playing the new game and the losers are not. Make the future split between a clearly negative outcome and a clearly positive one. Raskin's analogy is Star Wars: Luke refuses Obi-Wan's call to adventure until the aunt and uncle die. Until then, the future feels okay enough.
3. Name the object of the new game. The rallying cry. Zuora: "turn customers into subscribers." Airbnb: "live anywhere." 360Learning: "upskill from within." Often phrased as a question that brings the listener in as a co-adventurer.
4. Name the obstacles. What used to be called "problems we solve" become obstacles to a meaningful goal already positioned as life and death. Same content, different emotional weight.
5. The magic gifts. Now you talk about your product. Raskin's line: "You're not pitching product. Your product is like a prop for making the story come true."
What does this discipline look like in 2026, when the same CEO has AI tools that can spray ten variations of the narrative across every channel by lunchtime?
Three AI-era failure modes
I have watched each of these play out with clients in the last twelve months.
The Drift. AI personalizes the message per audience until the audiences no longer recognise they are getting the same story. Sales hears a version about closed-won ARR. Engineering hears a version about velocity. The board hears a version about EBIT. The middle management layer is the most disengaged it has been in a decade, and that is the layer where the CEO's narrative has to land. When it lands as fourteen different stories, it does not.
The Spray. AI generates infinite comms artifacts. CEO videos, Slack posts, town hall decks, customer emails. Quantity overwhelms coherence. A Stanford study found 41% of workers encounter AI-generated workslop monthly, costing about two hours of rework per instance and creating a $186 invisible tax per worker per month. The narrative is buried under its own volume.
The Outsource. Sebastian Siemiatkowski at Klarna delivered Q1 2025 earnings via an AI-generated avatar of himself. By May 2025 Klarna was rehiring human customer service agents, and Siemiatkowski admitted publicly: "we went too far." The avatar was the visible artifact of a deeper move Klarna had to walk back. Authoring had been outsourced, and the conviction drained out of it.
The cure has the same shape Raskin's framework already has, with one AI-era addition: amplify, do not generate. Both halves of that pair matter.
- Refuse to amplify and you cede the bandwidth war to leaders who are reaching forty markets while you reach two.
- Hand AI the authoring and you flatten what made the bandwidth worth fighting for.
Tobi Lütke's April 2025 AI-first memo at Shopify is the cleanest example of this discipline. The memo travelled from his desk to every employee and journalist who saw it with the same sentences intact, partly because he wrote it himself and partly because nobody softened it on the way out. Whether you agree with the policy or not, the narrative did not drift. Satya Nadella is doing the same work in real time this quarter, publicly fighting the "AI slop" framing and trying to author "bicycles for the mind" as the counter-narrative in his own voice.
This is the same discipline I unpack in this issue's leadership section on bold direction and AI. Megaphone, not author.
The diagnostic
One question to run on your own AI transformation this week.
Can a frontline employee in your organization state your AI transformation thesis in your CEO's own words within ten seconds?
If yes, you are already in the 5%. If not, the work is upstream of the next AI tool you were planning to buy.
Your action step
Block ninety minutes this week. Open a single page. Write out Raskin's five elements for your AI transformation in your own words, without AI assistance. Name the shift in one sentence. Name the stakes in two outcomes. State the rallying cry as a question. List three obstacles your transformation removes. Describe how your AI investment is the prop that makes the story come true, not the story itself.
Read it back the next morning. If it sounds like you, it has a chance of landing. If it sounds like a CEO, start again. The page is not the deliverable. The voice is.
For a deeper dive, watch Andy Raskin's episode on Lenny's podcast. The strategic narrative framework is laid out there with examples and stress tests.
This is the work I do with leadership teams in AI strategy advisory engagements and in working backwards sessions when an AI transformation has a portfolio but no story. If your AI program is in the 94% and you want to be in the 6%, that is usually where the fix lives.