AI Manifesto: Stage 3, Commit
Write the Rules Before You Buy the Licences
Most organisations won’t fail at AI because the model is weak. They fail because the organisation never made a decision, in daylight, about what it will and won’t allow AI to do.
That’s why Stage 3 of the Human-First AI Adoption Framework is called Commit, not “Choose a vendor”. The question is almost disarmingly basic: “Do we have the leadership, budget, and ethics to change how we work?”
The Commit stage exists to force decisions before dependency. Before licences are purchased. Before workflows quietly shift. Before “temporary” tools become permanent operating habits.
The framework defines commitment in practice terms. Visible executive sponsorship, because AI adoption without it stalls fast. An ethics position that forces the awkward conversation: what will we never use AI to do. A human-in-the-loop policy agreed before tools are deployed. Budget for training and change management, not just licences.
And then there is the guardrail that matters most: publish your AI principles internally before deploying anything.
That single act changes AI from technology initiative into an organisation position.
This is where the manifesto reiterates that it is not simply a poster. “Responsibility Has A Name” is comforting until there is an incident, a complaint, or a regulator asking who approved the system’s use. “Humans Decide. AI Assists.” sounds reassuring until your organisation quietly treats AI output as the final word. Commitment is where you prevent that drift by making the rules explicit before the tools become normal.
The Amazon Lesson: Clarity Before Build
If you want a useful comparison, look at how Amazon institutionalised clarity before build. The “Working Backwards” approach is famous for forcing teams to write a press release and FAQ before product development begins, so the organisation is aligned on the customer problem and intended outcomes before anyone ships capability. That isn’t a gimmick. It’s a discipline designed to stop the most common corporate mistake: building something impressive and discovering later that nobody can explain what it’s for.
Commit applies the same logic to AI. Your internal AI principles are the “press release” for your own people. They answer: what this is for, what is off-limits, who owns it, and how it will be governed. If those questions can’t be answered, the organisation is not ready for scale, even if it is ready for experimentation.
Strategy Is the Choices You Defend When the Pressure Rises
Strategy also shows up here, in a way that AI programmes often dodge. Michael Porter’s point that strategy involves choices and trade-offs is hard to argue with when you’re staring at a technology that can be used everywhere. In practice, Commit is where you set the boundaries you will defend even when the pressure rises: no automation of empathy or ethical discretion, no unowned systems, no “we’ll fix it later” on data discipline.
“Strategy is about making choices, trade-offs; it’s about deliberately choosing to be different.” — Michael Porter, “What is Strategy?” (Harvard Business Review, 1996)
Deploying AI faster than competitors may improve operational effectiveness. It does not automatically create strategy.
That’s not anti-AI. It’s how serious industries behave when tools become powerful enough to introduce systemic risk. Financial services didn’t wait for the first trading scandal before introducing controls; it learned the hard way that controls are part of the product. Aviation doesn’t treat checklists as optional because the aircraft flew fine yesterday. It treats procedure as the price of operating at scale.
The Budget Line That Most Organisations Get Wrong
The Commit stage also tackles the bit that budgets consistently underestimate: training and change management. This matters because AI adoption is a behaviour change, not a software deployment. If your staff don’t know when not to use AI, or how to verify outputs, the organisation will either under-use tools or over-trust them. Both are expensive. And both are predictable.
Request the full AI Manifesto
What a Strong Commit Stage Actually Produces
A strong Commit stage produces artefacts that feel slightly uncomfortable because they are specific: an internally published principles document. A formal human-in-the-loop policy. A named executive sponsor. A budget line that includes education, communication, and governance alongside licensing cost.
They are the foundations that make the later stages possible.
Upskill becomes targeted because the policy defines what people must be able to do. Pilot becomes safer because ownership is non-negotiable. Governance becomes credible because the organisation chose its limits before the technology chose them instead.
Growing the Decision-Making Spine
The AI race rewards speed. But sustainable adoption requires something different: a decision-making spine.
Commit is where you grow it.
Request the full AI Manifesto
