Write your company's AI manifesto (with a template)

An AI manifesto is a short, public statement of how your company will and won't use AI. Not the rules, the stance: why you use AI at all, what you value about it, the bright lines you won't cross, and who stays accountable for the decisions. It's deliberately the why, where your acceptable-use policy is the how. The two work together, and conflating them is why so many companies have a policy nobody likes and no stated values to explain it. A manifesto people can read in one sitting sets the tone the policy then operationalizes, which is exactly the move our guide on how to set an AI policy picks up. Think of the manifesto as the layer that says what kind of company you intend to be with AI, before any clause tells someone what they can and can't paste into a chatbot.

The reason to write one isn't ceremony. It's that the alternative, saying nothing, is quietly costing you.

Silence is the expensive default, and the data shows it. Pew found 52% of US workers are worried about AI in the workplace, with far more anxious than hopeful, and that fear has consequences. Ivanti found about 30% of employees secretly use AI tools because they worry their role could be cut, and Deloitte's trust index, reported by Harvard Business Review, recorded a sharp drop in employees' trust in company-provided AI over a single stretch of 2025, as people grew uneasy with technology taking over decisions that were once theirs. When people don't know where the company stands, they hide what they're doing and trust craters. A two-paragraph statement that says, plainly, we use AI to level you up, not to watch you, and a human stays accountable, does work no policy clause does. That's why a manifesto belongs at the front of any AI adoption effort, not as an afterthought.

It helps to read a few real ones. The Marketing AI Institute's manifesto states plainly that AI should be assistive, not autonomous, and that a human stays in the loop in all applications, and it frames AI as a way to free people for creativity and judgment, which is close to the right tone. Syngenta's leads with human oversight and even includes an employee-empowerment commitment, though it reads more like a governance document than a human one and stays generic enough to belong to any company. Microsoft and Google each publish a short set of responsible-AI principles, and the length lesson is instructive: six words you can remember beats twelve you can't. The pattern across the good ones is short, specific to the company, and human in tone. A manifesto people can't recite isn't doing its job, so write yours to be remembered, not filed.

The sharpest counterargument is that manifestos are virtue-signaling: there are already seventy-plus sets of AI principles out there, most of them ethics-washing that never touches the product roadmap, and nobody reads a values page, so just write the policy and skip the poetry. The critique is right about bad manifestos, and that's the useful part. A manifesto is fluff only if it floats free of consequences. The fix isn't to skip it; it's to keep it short, name what you won't do, which a policy alone rarely does, and wire each principle to something operational. And the fear data is the case for writing it, not against: when people don't know the stance, 30% hide their AI use and trust falls, so a clear, short statement is the cheap fix for an expensive problem. Concede the critique honestly, then answer it the only way that holds, with a manifesto that's short, specific, paired with the policy, and revisited on a date. Whoever holds the Head of AI mandate is usually the right owner to keep it honest, as part of the broader AI transformation.