I gave a keynote last week, and I asked the room how many people had an AI manifesto for their organization. Two or three hands went up out of maybe a hundred people.

That's a problem, but it's not the problem I want to talk about. The bigger problem is that most of the AI manifestos I've seen - including a few we've helped clients write - are so vague they could justify any behavior. “We will preserve our human ingenuity while embracing AI.” What does that actually mean? What does an employee do differently on Monday morning because that sentence exists?

Nothing. They do nothing differently. That's the issue.

The manifesto most companies write vs. the one they need

Most organizations that do create an AI manifesto treat it like a marketing document. It's glossy, it's aspirational, and it communicates a vibe: we're all in this together, AI is an opportunity not a threat, we value human creativity. 

Employees read it once, nod, and never think about it again. It's not something you'd ever reference in a performance review because there's nothing concrete enough to hold someone to.

The manifesto your organization actually needs is different. It should be specific enough that you can look at any employee's behavior and say: you're doing this, or you're not.

What a useful manifesto actually contains

The first line of ours communicates that Section is becoming a supercompany. That's the premise. Everything else has to support it. If a line item in your manifesto doesn't clearly connect to the organizational goal, it shouldn't be in there.

Beyond the premise, your manifesto needs concrete commitments - what we've been calling commandments. Not ten aspirational principles. Ten things you actually do and can be held accountable for. A few examples of what I think belongs:

Proficiency targets that are specific and measurable. We track AI proficiency across the organization - novice, experimenter, practitioner, expert. The expectation at Section is that 100% of the company operates at practitioner or expert level. That's not a vibe. It's a number we measure and a standard we hold people to.

How much time people should dedicate to AI. I genuinely believe every person in a company needs a minimum two-hour block on their calendar every week for AI experimentation. Not "use AI when you get a chance." A blocked, protected two hours. I'll be checking whether people at Section are actually doing this - and asking their managers why if they're not.

‍How AI expectations will be enforced. At Section, AI proficiency will be part of your performance review. If you're a practitioner, you should be on a path to becoming an expert. If you're an experimenter, that's a conversation your manager needs to be having with you now, not in six months.

What you expect from managers specifically. This is where most manifestos fall apart. They set expectations for ICs but say nothing about what managers need to do. Managers should be AI power users, and they should be actively enabling their teams' experimentation, not passively allowing it.

How you think about hiring and resourcing. At Section, two of our commandments are: Exhaust AI before asking for new resources, and don’t hire anyone who doesn’t use AI or only uses it occasionally. The latter is a hard gate on any new candidates that’s easy to enforce.

Make it a living document

The manifesto you write today will be outdated in three months. What “AI proficient” means is changing constantly as the tools evolve. The proficiency bar that was ambitious in January is table stakes by June.

My recommendation: revisit your manifesto every quarter. Not a full rewrite - but a real review of whether the expectations still make sense and whether the targets are still ambitious enough. Reference it in company meetings with actual data. Here's where our proficiency levels are. Here's how many automations were built this month. Here's where we're falling short.

If your manifesto isn't showing up in performance conversations and company-wide updates, it's decoration. And you don't need decoration. You need a document that actually changes behavior.