This post was originally published by Saanya on her Substack, 'The Change Constant' – subscribe HERE!

When Tobi Lütke’s Reflexive AI memo leaked recently, it did what good memos should: set the tone – and split the room. Some hailed it as a rare show of CEO-level urgency. Others thought it was heavy-handed. But whether you loved it or not, the memo is a cultural artifact. It captures the very real tension every enterprise leader is wrestling with: How do you take a 8,000+ person org on an AI journey when the terrain is shifting daily?

Because here’s the reality playing out inside every big company right now: AI is flooding the enterprise like water into a leaky basement. It’s coming from the top (executive mandates), the bottom (rogue custom bots) and sideways (vendor partnerships your CTO heard about on a panel). Everyone’s experimenting. No one’s aligned. The result? Teams are duplicating efforts with no shared standards, security teams are stuck in reaction mode, finance is seeing shadow AI spend across 20 cost centers and legal is panicking.

I’ve had a front-row seat to this lately. Every conversation with an enterprise company sounds the same: “How do we move fast enough to feel AI-native, without buying into hype or derailing our core workflows?”

AI in the enterprise isn’t just about the tech. It’s about threading a needle between innovation and structure. Executives aren’t losing sleep over model benchmarks. The questions they are grappling with are brutally tactical:

• How prescriptive should I be? Should we recommend AI tools in every vertical, or let teams self-serve?
• How secure are we, really? What guardrails do we need to protect data without blocking progress?
• What’s the line between innovation and chaos? Everyone’s building bots – but who’s maintaining them? Who owns the outputs?

This is the less glamorous side of AI adoption. It’s not about model architecture. It’s about governance, incentives and behavior change at scale.

Some playbooks that actually work:

• Run a hackathon. Not for PR - for bottom-up discovery. Let employees build, showcase, and vote. Then standardize on the winners. It fosters inclusion without compromising control. And it upskills the org in the process.
• Watch what spreads. Watch what spreads organically – then sign an enterprise contract after usage tips. Flexibility early, scale later.
• Set the tone from the top. This is where Tobi’s memo shines. It’s not a technical brief. It’s a cultural directive. It signals urgency. It clarifies expectations. And it gives permission to move. Most employees aren’t waiting on tech – they’re waiting on clarity.
• Invest in internal infrastructure. Prompt libraries. Guardrails. LLMOps. The invisible scaffolding that makes experimentation safe, repeatable and useful. Right now, 90% of enterprise AI is built on duct tape and vibes. That’s fine for prototypes. It’s fatal for production.

AI may be dynamic and fast. But organizational change is slow, human, and political. That tension isn’t new – it’s just wearing a new outfit.

The real winners won’t be the most “AI-native.” They’ll be the most change-literate. The ones that can absorb fast-moving tech without imploding structurally.

So yes, AI is sexy. But under the hood? It's the same old grind: incentives, governance, coordination, cultural permission. The real work starts where the headlines stop.

Tech changes fast. Humans don’t.

If you can manage that delta, you win.

Find the full memo here.