Patent strategy is typically treated as a drafting problem. An inventor submits a disclosure. A drafter produces claims. A searcher checks prior art. Then the claims get rewritten.

This sequence has a structural flaw: the most consequential decisions — what to claim, at what scope, with which structural boundaries — are made implicitly, under time pressure, after significant resources have already been committed.

The strategic question what should we claim and why? never gets answered on its own terms. It gets absorbed into the drafting process, which is designed to produce text, not to validate direction.

The result is claims that reflect what was written first, not what was worth protecting most.

Generative AI has made it significantly easier to produce patent text. A claim set that once took hours to draft can now be generated in minutes from a disclosure summary.

This is a real reduction in friction. But it does not address the underlying design problem — it accelerates it.

Generating claims faster from a poorly chosen direction produces more output at lower quality. The speed advantage of AI is only realized when the strategic input is sound: when you know which direction to draft toward, why that direction is defensible, and how the claim structure should be organized before the first sentence is written.

The innovation space for any given disclosure is not a single point. It is a landscape — with multiple possible directions, each carrying different prior-art risk, structural strength, and scope potential. Exploring that landscape systematically, before committing to a draft, is a design problem that requires methodology, not just generation.

AI creates the opportunity to explore this landscape at scale. But the exploration still requires structure — and that is where most tools stop short.