The SoW that felt like control
A buyer we will call Steve used to wait for his suppliers to draft the Statement of Work. It was slow, and the scope was at least grounded in what the supplier knew how to deliver. Then he started drafting it himself with generative AI, and the output was remarkable: workstreams he had not considered, governance structures, risk frameworks, acceptance criteria. For the first time he felt in control of defining the work.
He was not. He was reading what a language model had invented from statistical patterns, and because it was articulate and thorough it became the basis for the engagement. The supplier priced against it. The plan followed it. Six months later Steve was running a programme twice the size of the problem he had set out to solve.
The instinct generative AI rewards here is customisation, and customisation is the wrong instinct for a commercial document. Every generation a model produces is a fresh probabilistic walk, statistically unique by design. For a buyer, that means each engagement now starts from a scope that has never been reviewed, never been priced, and never been tested against delivery. The fluency is the trap. MIT's Media Lab measured brain activity while people delegated writing to a model and found the weakest engagement of any group, an effect they named cognitive debt that persisted after the tool was taken away. Microsoft Research found the matching pattern: the more confident the user is in the output, the less they scrutinise it.
Treating an infinite document factory as control is the opposite of control. A buyer who can regenerate forever has no basis for choosing between drafts except which one reads best, and reading well and scoping accurately have never been the same skill.
AI has a real place in scope documentation. It is after the structure, not before it: scanning a document that already exists, extracting obligations, scoring risk. Generation belongs nowhere near the blank page.
The instinct behind most weak SoWs is to write each one fresh. Every freshly authored scope is a novel document that no one has reviewed, priced, or tested against delivery, and that novelty carries a commercial cost the buyer rarely sees until it arrives. The standard inverts the instinct. It begins from a baseline the enterprise has already approved, eleven sections and fifty-one elements, and permits variation only where the work genuinely differs. The schema turns that structure into data, the substrate AI needs to extract obligations, score risk, and govern a portfolio, rather than generate one more unreviewed draft. AI reads, scores, and governs a SoW that already exists; it does not invent the one nobody has reviewed.
The launch
statementofwork.org launches on 10 August at the World Commerce and Contracting APAC Summit in Sydney. The full standard and engagement types, and the first learning and certification modules go live the same day. You signed up before any of it was public, which means you see it first.