What is the 30% rule for AI?

The 30% rule is an informal principle: let AI do the bulk of a task, roughly 70%, and keep a human on the final 30%, the judgement, verification, context and accountability. It is not an official standard, and versions vary, but the idea underneath is sound and usefully blunt: the machine drafts, a person owns.

You will meet the rule in a few costumes. Sometimes it is the split above. Sometimes it is quoted as AI can automate about 30% of most jobs, an observation about tasks rather than quality. Occasionally it appears as an adoption threshold. The versions share one insight worth keeping: AI changes the shape of work rather than replacing whole roles, and the human contribution concentrates at the ends, framing the task well going in, and applying judgement coming out.

The practical use of the rule is as a design question, asked per task: which parts of this job can the tool carry, and which parts must a person hold? For a proposal, the machine drafts structure and prose while the person owns the pricing, the promises and the client knowledge. For meeting notes, the machine transcribes and summarises while the person confirms the decisions and actions. Run that question across a role and you get something better than a ratio: a written split, which is also the beginning of sensible training and sensible policy.

The honest caveat: the rule fails when the 30% is skipped rather than kept, and skipping is tempting precisely because the 70% arrives looking finished. Confident, polished and wrong is the failure mode of this whole technology, and the human share exists to catch it. Treat the 30% as the part of the job you are actually paid for, and the rule serves you well. If you want help designing the split for your team's real tasks, call 1800 456 567.