AI transformation theater: What CEOs must do differently

Ahead of SuperReturn Operating Partners, Emilia Sherifova, Founder and CEO of LegionX, Inc, paints a picture. Picture this. It’s Q4 2025. A CEO walks into a board meeting with a deck titled “AI Strategy Update.” The deck is tight. The optics are perfect. Slide after slide shows “progress”: 15 AI tools deployed, 3 pilots underway, a Chief AI Officer hired, an “AI Centre of Excellence” established, consultants engaged. The board nods approvingly.

Six months later, the same CEO is back in the same room. One director leans in and asks, “We spent $5.7M on AI last year. What actually changed?”

Silence.

The tools are there, but adoption is stuck at 8%. The pilots generated plenty of reports, but zero results. The Centre of Excellence became a place where ideas go to get polished, not shipped. And the most important part hasn’t budged: nothing about how the business operates has changed.

The pattern is obvious once you see it: the CEO didn’t own it. Not as a top-three priority. Not as the single point of accountability to the board. Not as the person narrating what’s changing and what’s not. Not as the person willing to force adoption when it gets uncomfortable. And in that vacuum, the organisation does what it always does: it buys tools, launches pilots, forms committees, ships decks. It looks like progress. It is not. Because AI isn’t a “tech initiative.” It’s a change in how business operates. And only the CEO can break the tie when incentives, processes, priorities, decision rights, and internal politics collide. We’ve seen this many times over. It’s a transformation theatre, not by design, but because real transformation is arduous.

Back in 2010, it was: “We’re moving to the Cloud!” The result, in most cases, was lift-and-shift migrations, the same mess on a new server, a bigger AWS bill, and somehow even more people.

Then 2015: “We’re going SaaS!” The result: disjointed tools that don’t talk to each other. And instead of fewer people, we hired more just to manage integrations.

Then 2024: “We’re implementing AI!” Same story. Chatbots. Pilots. Steering committees. And boards are starting to ask uncomfortable questions.

The pattern is always the same: hype, massive budgets, consultants, decks. And a year later, nothing fundamental has changed about how work gets done.

The uncomfortable truths

Here’s what I learned from the transformations that actually worked, across financial services, from high-volume trading systems to large enterprise operating environments.

First: “My CTO is handling AI” is how this turns into theatre. Tech builds capability. The business continues to operate in the old way. Sometimes that’s because incentives and workflows never change. Sometimes it’s because tech built something the business didn’t ask for and won’t use. AI works only when business and tech co-own the workflow, the build, and the adoption. Otherwise, it’s just software that looks impressive and changes nothing.

Second: “We’re piloting” is code for “we’re avoiding hard decisions.” Pilots are where transformation goes to die because they let you postpone the real work: reassigning your best people, changing how performance is measured, and making operational tradeoffs that hurt in the short term.

Third: if adoption is optional, it won’t happen. Meta just made AI usage mandatory in 2026 performance reviews. Amazon’s CEO told employees that AI will eliminate roles, learn it or leave. Coinbase’s CEO personally followed up with engineers not using AI tools and fired those who refused. That isn’t theatre. That’s what forcing adoption actually looks like. Meanwhile, most companies “encourage” AI usage. They run lunch-and-learns. They send newsletters. They hope people volunteer. Then they get 8% adoption and wonder why the transformation failed.

The problem nobody’s talking about

Here’s what keeps me up at night: AI moves weekly, but companies move annually.

I grew up in a world of Soviet-style five-year plans, rigid, paper-heavy, and obsolete the moment they were printed because the world moved faster than the central planners. Today’s annual AI strategy is no different. If you spec a solution in January based on what models could do, it’s a museum piece by Q3. By July, capabilities have jumped 10x and costs have collapsed, but you’re locked into an architecture designed for last year’s AI.

Annual planning cycles guarantee you’re building for a world that no longer exists. In this era, the only durable advantage is your agility and cycle time: the speed at which you can learn, rebuild, and ship.

What actually works

After watching this fail repeatedly, and leading a few transformations that actually worked, here’s what separates real change from theatre.

The first is surprisingly simple: run AI like you run sales. Weekly 30-minute CEO reviews. Demo what shipped, no slides allowed. Track adoption metrics. Make one decision to unblock progress. If you won’t personally chair this meeting every week, don’t fund the program. It will drift into storytelling.

Second: treat AI like a business continuity risk, and give the board something real to govern. Not a narrative or a “tool count.” A one-page quarterly scorecard that covers outcomes, adoption, and risk. And then insist on two questions that are hard to bullshit:

• “Show me one person whose job fundamentally changed because of AI, and walk me through exactly how.”
• “What did we stop doing because AI does it now?”

If leadership can’t answer those with specifics, you’re funding theater because you’re not subtracting work. You’re just layering AI on top of it.

And recruit at least one AI-native board member, or bring in an AI-native board advisor for two quarters. Your board can’t interrogate risk or progress if nobody in the room has shipped this in the real world.

Third: business and tech must co-own outcomes. Not hand-offs. Co-ownership. Both were measured on the same metrics. If either side can say “that’s not my problem,” you’re doing theater.

Fourth: build a funding engine powered by efficiency and growth, not headcount cuts. Start with small bets that can pay for the next one. Run 60-day pushes focused on one workflow where you can remove friction, stop doing the same work twice, and free up real time and dollars, then reinvest those gains into the next wave. Pay bonuses only when it’s live in production, the team is actually using it (60%+ adoption), and the impact shows up in the numbers. And if customers will pay for the outcome, pre-sell premium service tiers to help finance the build. You can fund this big if you have the budget and the conviction. In fact, sometimes you should, especially when you know the workflows you’re going after, and you need speed. But most CEOs don’t get handed a clean “AI budget.” They have to manufacture it. The goal isn’t to avoid spending. It’s to finance the work in ways that don’t destroy morale, don’t gut the core business, and don’t require a heroic leap of faith.

And then there’s a rule I’ve become almost religious about: the 90-Day Rule. If it isn’t shipped, adopted, and showing in the financials within the first three months, stop funding it.

The fork

There were Walmarts, and there were K-Marts.

There is a new class of AI-native platforms emerging: firms rebuilding operations around AI as the operating system from Day 1. They aren’t waiting for certainty; they are creating compounding advantage while incumbents “pilot.”

Every CEO faces the same choice. Be the Founder of the new system: reinvent, build AI-native capability, and lead your people into the future. Or be the Caretaker of the old one: protect legacy processes, move slower than competitors, and watch margins and talent erode.

The question isn’t whether you’ll use AI. It’s whether you’ll use it before it’s too late to matter.

Most CEOs don’t have an answer for their board yet. Do you?

Emilia Sherifova is the CEO and Founder of LegionX, an AI-native platform for asset management and regulated enterprises. She is the former CTO of Northwestern Mutual and a former Partner and Chief Information and Innovation Officer at KKR.