Innovation Theater at Scale: When AI Mandates Create Cacophony

You got the memo. “Be a builder.” “Reflexive AI usage.” “Demonstrate AI can’t do this before requesting headcount.” Maybe it came with a link to a copilot license and a Slack Support channel. What it didn’t come with was a list of what to build, who owns the outcome, or how anyone will know it worked.

Six months later, three teams are racing each other to ship a meeting summarizer. The email-thread digest exists in twelve flavors. Two of your strongest engineers are quietly demoralized because their version of the customer-churn predictor lost a competition nobody told them they were in. The dashboard shows record AI activity. The P&L shows nothing.

When a company tells everyone to build with AI but won’t say what to build or why, it has not empowered its workforce. It has outsourced its strategy to a thousand uncoordinated contractors who happen to share a payroll.

Here’s what’s inside this piece:

Here’s how the tune goes. A Fortune 500 leadership team reads the same three McKinsey reports. They watch their competitor announce an “AI-first” reorg. They feel the board’s eyes on them. So they issue the directive: every team will integrate AI into their workflow by end of quarter.

What happens next is predictable. Marketing builds a copy generator. Legal builds a contract summarizer. Sales builds a call-notes bot. Finance builds a forecast assistant. HR builds a resume screener. Engineering builds an internal Copilot wrapper. Twelve teams. Twelve models. Twelve vendor contracts. Zero shared infrastructure. Zero shared data layer. Zero shared definition of what “good” means.

A 2024 BCG survey of 1,400 executives found that only 26% of companies running AI initiatives had moved past proofs of concept into measurable value. The 2025 MIT State of AI in Business report sharpened the picture: roughly 5% of pilots produced rapid revenue acceleration; the rest delivered little measurable P&L impact. The MIT authors call the gap a “learning gap” between tool and workflow. I’d call it something simpler — a framing failure. The 74% didn’t pick the wrong AI tools. They didn’t pick at all. “Use AI” isn’t a strategy; it’s a posture, and postures don’t show up in the P&L.

Scale this posture across the whole company and the bill compounds and the noise gets louder. You haven’t just wasted a quarter — you’ve fragmented the productivity and agility that would have let you make a real bet later. The cost of a blanket mandate isn’t the wasted compute. It’s that when a competitor makes a real bet — rebuilding their entire customer-service motion around a single agentic system trained on proprietary data — you can’t respond. Your AI investment is scattered across dozens of pilots that don’t share a backbone, a data layer, or a definition of success. You don’t have one bet to double down on. You have dozens of half-bets to defend. If the cost is invisible until a competitor forces the issue, you need a way to spot it earlier. Here’s what to look for.

If three or more of these are true on your team right now, you are not transforming. You are performing.

One. The KPI is “AI adoption,” not a business outcome. Leadership tracks how many teams have “integrated AI.” Nobody tracks whether revenue per employee, gross margin, time-to-resolution, or any actual P&L line moved. Adoption is an input. The mandate has confused it for an output.

Two. There’s no shared data layer. Each team’s AI work runs on its own copy of customer data, its own embeddings, its own eval set. When the marketing copy bot and the sales follow-up bot describe the same product differently to the same customer, nobody owns the contradiction.

Three. The “AI council” meets monthly and decides nothing. It approves vendor contracts. It rubber-stamps pilots. It does not kill projects. A governance body that has never killed a project is not governing — it’s catering.

Four. Senior leaders cannot finish the sentence “We are using AI to become the company that ___.” If the CEO, the CFO, and the head of product give three different endings, you don’t have a strategy. You have a vibe.

Stripe didn’t tell every team to “leverage AI.” They identified that fraud detection and developer-facing documentation were the two surfaces where probabilistic systems had the highest leverage on their core economics, and they concentrated investment there. Klarna didn’t run forty pilots. They rebuilt customer service around a single agent and reported, in a 2024 earnings update, that it was doing the work of 700 human agents and resolving issues two minutes faster. Argue with the headcount math if you want — the shape is what matters: one bet, one surface, one number on the wall.

Strategic direction sounds boring compared to a mandate. It sounds like:

That’s a sentence that kills projects. A sentence that kills projects is the only kind of strategy that’s real.

Here’s the artifact. Screenshot it, paste it in a doc — whatever gets you to actually do it. Run it for your own team first, then bring the filled version to your skip-level.

Question

Your answer

What is the one business outcome our AI work must move?

Reduce cost-to-serve in mid-market support by 30% by Q4

A single, numeric, P&L-linked outcome

Which AI projects are funded against that outcome?

Agent-assisted ticket resolution; auto-classification; macro generation

Three or fewer, all defensibly tied to the outcome

Which AI projects are funded but NOT against that outcome?

Sales-deck generator; internal-wiki Q&A bot

This list should make you uncomfortable

What did we kill in the last 90 days?

Killed: marketing-copy generator, replaced with vendor tool. Killed: forecasting pilot, no signal

If the answer is “nothing,” your council is theater

Where does the data for the funded projects live?

Unified support-events warehouse, owned by data platform team

One owner, one source, one eval set

Who can be challenged if the outcome isn’t met?

VP of Customer Operations

A specific role/person. If your answer is a committee, the committee is the problem.

If the right column makes you flinch, that’s the audit working. The flinch is the strategy showing up.

When someone proposes a new AI project on your team, ask this and watch the room.

A real strategic project has a clean answer to both halves. A theater project has an answer to the first half and a blank stare on the second. The blank stare is the indicator. Every choice has a cost. If a project has no cost — no thing it makes harder, no road it closes off — it isn’t a strategic move. It’s a side quest with a budget.

The companies that come out of this AI cycle ahead won’t be the ones that mandated the most. They’ll be the ones whose leaders had the nerve to say what the company was for, what it wasn’t for, and which AI bets earned the right to exist because they served the first thing and ignored the second.

A mandate without a thesis is not a strategy. It’s a thousand contractors on the same payroll, building in a thousand directions, calling it transformation because nobody at the top was willing to choose.

Choose.