A slide deck with three pillars and a rocket graphic is not a strategy. Here's what actually separates the two, and why most companies stall out somewhere between them.
Most AI strategies are documents people quote, not documents that decide anything. What separates a real one from a for-show one comes down to three things: one named person with actual authority and a real budget, a short list of three to five priority use cases each tied to a specific measurable outcome, and a working process that reviews, on a set schedule, what gets scaled and what gets killed. Miss any one of the three and the other two tend to become decoration. The fix isn’t a full rewrite. Most companies can convert their existing slide deck into a real plan in a single working session by naming an owner, attaching a number, and writing the kill criteria before the budget gets approved rather than after a pilot underperforms.
Picture the all-hands meeting. Slide twelve is the AI strategy: three pillars in a row, a roadmap graphic with a little rocket climbing toward the corner, the word “transformation” appearing twice. The room nods. Somebody claps. Three weeks later, nobody in that room could tell you which of the three pillars anyone is actually accountable for, what the budget line looks like, or what would have to happen for one of last year’s AI projects to get killed instead of quietly renewed.
That gap, between the slide and the actual decision-making underneath it, is what this piece is about. It’s also more common than most leadership teams are comfortable admitting out loud, though plenty admit it quietly, in the way people admit things when nobody’s taking minutes.
Writer, a company that sells enterprise AI software, surveyed 2,400 knowledge workers between December 2025 and January 2026, including 1,200 C-suite executives, working with the research firm Workplace Intelligence. Worth saying upfront: this is a vendor-commissioned survey, not neutral academic research, and Writer has an obvious interest in AI adoption looking urgent. Even so, the finding is hard to wave away. 75% of the C-suite executives surveyed said their own company’s AI strategy is “more for show” than for actual internal guidance.[1] Almost four in ten said they don’t have a formal plan to use AI tools to drive revenue at all.[1]
That’s self-reported, by the people who wrote or approved the strategy in question.
| # | Question |
|---|---|
| 1 | Is there one named person who could actually be held accountable if this strategy fails? |
| 2 | Does that person control a real budget line for it, separate from general IT or software spend? |
| 3 | Are there fewer than five priority use cases, chosen on purpose, rather than one from every department that asked? |
| 4 | Is each priority use case tied to a specific number it’s supposed to move, and by when? |
| 5 | Has your company actually killed or paused an AI initiative in the past year because it wasn’t working? |
| 6 | Could a new hire read the document and know what your company has decided NOT to do with AI, not just what it plans to do? |
Score one point for every “yes.” 5 to 6: a real strategy, actively guiding decisions. 3 to 4: parts of the machinery exist, but ownership or follow-through is missing somewhere. 0 to 2: a communications document wearing a strategy’s clothes.
If no single person could be fired over this strategy failing, it’s not a strategy: it’s a slide deck with a title.
None of this means the leadership team is lying. Most for-show strategies start as a genuine attempt to get everyone pointed in the same direction. The problem shows up later, when the document stops being revisited and starts being cited: quoted in board meetings, pasted into a pitch deck, without anyone checking whether the decisions underneath it actually changed.
Say a CFO at a 400-person insurance brokerage pulls up the company’s “AI Strategy 2026” PDF the night before a board meeting. Fourteen pages. A vision statement. A maturity-model graphic that looks suspiciously similar to one she’s seen on a vendor’s website. What isn’t in it, anywhere: a named person accountable for any of the four “strategic pillars,” a budget figure separate from the general technology line, or a sentence describing what happens if a pilot doesn’t work out. She closes the laptop and writes one line in her notes: “This is marketing, not a plan.”
That scene repeats itself, with different job titles and different pillar names, at a lot of companies that would consider themselves ahead of the curve on AI. A few patterns show up often enough to be worth naming directly.
Worth being precise about what this piece is not about. A written AI usage policy, what’s allowed in a tool, who reviews outputs before they reach a customer, is a real and useful document, and a different thing entirely from a strategy. Our piece on the AI governance gap covers that policy and oversight layer: the rules for how AI gets used safely day to day. This piece sits one level up, on whether anyone has actually decided which three or four AI bets the company is making this year, who’s accountable for them, and what happens if they don’t pay off. A company can have an airtight usage policy and still have a strategy that’s entirely for show, and the reverse is just as common.
It’s also worth separating this from plain adoption disappointment. The same Writer survey found 48% of executives calling their company’s overall AI adoption a “massive disappointment,” a number we unpacked in our piece on what the other 52% are doing differently, which is mostly about habits: training, workflow redesign, measuring the right thing. This piece asks an earlier question. Habits don’t fix an initiative that was never really prioritized, funded, or owned in the first place.
Strip away the slide design and the buzzwords, and a real AI strategy comes down to three things. Not five, not a maturity model with twelve boxes. Three, and all three have to be true at once, because any one missing quietly turns the other two into theater.
Not a working group. Not “the AI team,” a phrase that in most companies means three people from three different departments who each have a full-time job that isn’t this. A real strategy has one person whose performance review includes it, who can say no to a request that doesn’t fit, and who controls a budget line specific enough that finance could tell you the number without looking it up.
Three to five, not the eleven that show up when every department head gets to add one. Each one attached to a specific number: reduce average handling time by 15%, cut first-draft turnaround from four days to one, lift qualified pipeline by a stated amount. “Improve efficiency with AI” is not a use case. It’s a wish wearing a use case’s clothes.
This is the one that’s missing most often, because it’s the one nobody wants to write down before they’ve spent the budget. A real process names, in advance, what “working” looks like for each use case, then checks at a set interval, not whenever someone remembers, whether it’s still being used at all, whether that use survived past the first novelty month, and whether the actual metric it was funded to move has moved. A pilot nobody uses without being reminded, that never moved its number, has an answer: it gets killed, on a specific date, by a specific person, not left to quietly become part of the furniture.
| For-show version | Real version | |
|---|---|---|
| Owner | “The AI team” or “IT and the business,” nobody’s actual job | One named person, part of their performance review, with real authority to say no |
| Use cases | Seven to twelve pillars, one from every department, none ranked against each other | Three to five, each tied to a specific number and a deadline |
| Kill/scale process | Every pilot is permanently “in progress,” nothing has ever formally ended | A set review date, a stated bar for what counts as working, and a real record of what’s been paused or killed |
If your company’s strategy scores “for-show” on two of these three, the slide deck is doing more work than the plan underneath it.
If nothing has ever been killed, nothing was ever actually prioritized.
Use is not persistence. Persistence is not impact.
That third point is where those three matter more than a launch date, because “we tried it” and “it worked” are not the same claim. Plenty of AI pilots get used once, by the person who championed them, in the week after launch. Fewer are still being used a quarter later without a reminder email. Fewer still have actually moved the number they were funded to move. A real decision process checks all three before calling something a win.
Ask a leadership team who owns the AI strategy and watch what happens. In a lot of rooms, three or four people start to answer at once, which is itself the answer: nobody actually owns it, several people are adjacent to it, and the CEO assumes someone closer to the work has it handled.
The avoidance isn’t usually laziness. Naming one owner means naming who’s accountable when a well-funded pilot doesn’t pay off, and naming a budget means taking that money from something else with its own champion. It’s easier to leave it distributed, so no single name is attached to a number that might not move.
Gartner’s 2025 survey of 504 data and analytics leaders found that 70% of chief data and analytics officers already hold primary responsibility for building their organization’s AI strategy and operating model.[2] That’s a useful data point, not a universal rule. It describes what’s common at companies large enough to have a CDAO in the first place, which most mid-size companies aren’t.
Who the job actually suits depends heavily on company size and what’s already in place, which is worth being specific about rather than defaulting to “leadership” as a catch-all answer.
| Candidate owner | Best when | Watch out for |
|---|---|---|
| CEO or founder, directly | A small or mid-size company where AI decisions genuinely need to move fast and touch every function | Everything else on their plate quietly outranks it within a quarter, unless it’s on a calendar with the same seriousness as revenue reviews |
| COO | A company where AI use cases are mostly about operations: process, service, back office | Can default to efficiency use cases only, missing growth-side opportunities outside their usual remit |
| CDAO or Chief AI Officer | A larger organization with enough AI activity across departments to justify a dedicated seat and real headcount | Risks being seen as a technology role rather than a business one, which weakens their authority to say no to a business unit |
| A dedicated AI strategy lead, reporting to the CEO | A company that’s serious about this but not yet large enough to justify a full C-suite seat | Needs an explicit mandate from day one, or gets treated as a coordinator rather than a decision-maker |
None of these are wrong choices on their own. What’s wrong is leaving the seat empty, or splitting it three ways so nobody actually holds it.
Authority without a budget line is a title, not ownership.
Once a real owner exists with real authority and budget, the next question is usually how they organize the work underneath them: which parts get handled by a small set of specialized AI agents with named human owners of their own, rather than one general assistant everyone points at everything. That’s a different, more operational question, and we cover it directly in how to build an AI C-suite. Get ownership and priorities right first. The team structure underneath it is a second decision, not the first one.
None of this requires starting over. Most companies with a for-show strategy already have three or four of the right ideas buried in the deck. What’s missing is usually specific enough to fix in a single working session, not a quarter-long overhaul.
| What the slide says now | What to write instead |
|---|---|
| “Pillar 2: Democratize AI across the organization” | “Use case: AI-assisted claims triage in customer service” |
| (No owner listed) | “Owned by the customer service director, part of her Q3-Q4 goals” |
| (No budget listed) | “$40,000 tooling budget, 0.25 FTE from the ops analyst” |
| “Improve efficiency” | “Reduce average handling time from 9 minutes to 7.5 by end of Q2” |
| (No exit criteria) | “Review date: April 30. If handling time hasn’t moved by at least 10%, pause and reassess before renewing budget” |
Same initiative, rewritten with an owner, a number, a deadline and a way to actually stop. Do this for every pillar on the current slide before adding anything new.
Run that conversion across the current strategy and something useful tends to happen on its own: pillars that can’t survive being written this specifically get quietly dropped, because there was never a real use case underneath the language.
That’s the exercise working, not failing.
A short, practical starting sequence:
If your current strategy is mostly about scaling AI across the organization without naming which pilots earned that yet, it’s worth reading this alongside why AI pilots never scale, which walks through the operational reasons a pilot stalls even after leadership has already said yes.
None of this replaces a genuine planning session with the people who’ll actually own the work. A leadership team can run the conversion above internally and get real value from it. What’s harder to do alone is the part where an outside facilitator pressure-tests whether a use case is specific enough, or whether “the AI team” is quietly standing in for nobody, before the money’s spent. That’s the gap Future Factors’ Corporate Workshops and leadership sessions are built to close: working sessions built around a company’s actual AI priorities, that end with an owner’s name attached to a number instead of a pillar.
Start smaller than feels satisfying. Pick three use cases, name three owners, and put one review date on the calendar. That’s a smaller ask than rewriting the whole strategy, and it’s the difference between a document people quote and a document that actually decides anything.
It means the document exists, but it isn’t actually driving decisions. Nobody’s job depends on it, no budget is specifically attached to it, and it doesn’t say what the company would stop doing if an AI initiative wasn’t working. A 2026 survey of 2,400 executives and employees, run by Writer with the research firm Workplace Intelligence, found 75% of C-suite respondents describing their own company’s AI strategy this way: better for looking prepared to a board than for guiding what actually gets funded.
Three things, at minimum: one named person with real authority and a real budget line, a short list of three to five priority use cases each tied to a specific measurable outcome, and a working process that decides, on a set schedule, what gets scaled and what gets killed. Miss any one of the three and the other two tend to become decoration rather than a working plan.
It depends on size. At a smaller company, the CEO or COO can own it directly, as long as it’s protected on the calendar with the same seriousness as a revenue review. Larger organizations often formalize it in a CDAO or Chief AI Officer role; Gartner’s 2025 survey found 70% of chief data and analytics officers already hold primary responsibility for AI strategy at their companies. What matters more than the title is that one person, not a committee, can be asked in six months what changed because of it.
A policy is the rulebook: what employees can paste into a tool, what data is off-limits, who reviews an AI-assisted output before it reaches a customer. A strategy is a set of decisions: which two or three AI bets the company is actually funding this year, who owns each one, and what the exit looks like if it doesn’t pay off. A company can have a tight, well-run policy and still have a strategy that’s pure theater, and the reverse happens just as often.
Ask whoever’s supposed to own it one question: what was the last AI initiative the company actually killed, and why? A real strategy has an answer within a few seconds, complete with a reason. A for-show one gets a pause, then something about everything still being “early days” or “in progress.” That pause usually tells you everything you need to know.
The central claim, that a majority of C-suite executives privately view their own company’s AI strategy as more optics than substance, is checked directly against Writer’s own April 2026 press release for its 2026 AI Adoption in the Enterprise survey, run with the research firm Workplace Intelligence between December 2025 and January 2026 across 2,400 knowledge workers, including 1,200 C-suite executives. Writer sells enterprise AI software, so its survey is a vendor-commissioned data point with an obvious interest in AI adoption looking urgent, not neutral academic research, and that’s stated plainly in the piece rather than left implicit. The ownership statistic (70% of CDAOs hold primary responsibility for AI strategy) is checked against Gartner’s own May 2025 press release for its CDAO Agenda Survey. A separate, frequently repeated figure, that only 22% of companies have moved beyond proof-of-concept with AI and just 4% are creating substantial value from it, appeared consistently across several BCG-sourced summaries, but it could not be independently confirmed against BCG’s own page in the time available, so it was left out rather than included secondhand. The three-part definition of a real strategy (named owner, prioritized use cases, a working kill-or-scale process) and the ownership register are Future Factors’ own synthesis from working with corporate leadership teams on AI strategy, not a finding from either survey.
Weekly tips. Real examples. Practical help for busy professionals.
We care about your data, check out our privacy policy.