Guide
AI for Office Action responses: what it should and shouldn't do.
AI can take most of the reading, mapping and first drafting off an Office Action response. It can't take the judgment. Here is the line, and how ClaimMark-P handles each step.
What AI can do today
Read the action
Identify every rejection and objection (§101, §102, §103, §112, double patenting, claim objections) and keep the examiner's pin cites.
In ClaimMark-P: Imports the action from the USPTO by application, publication or patent number, or from an uploaded file, and parses every ground with its pin cites.
Map the art to the claims
Pull each cited reference and line up every claim limitation with the passage the examiner relies on.
In ClaimMark-P: Maps each claim element to the column and line of the cited reference, with a visible mapping confidence.
Propose options
Offer amendment and argument options for each rejection, each grounded in the specification and the reference text.
In ClaimMark-P: Proposes options per rejection; the attorney chooses the path.
Draft the response
Write claim amendments with specification support for every added limitation, and remarks that answer each ground.
In ClaimMark-P: Drafts amendments with support cited and remarks per ground, exported as an editable document.
What the attorney still decides
- • Whether the examiner's reading of each reference is right, after reading the passages yourself.
- • Which amendments are worth the scope they give up, given the client's product and business.
- • Whether every argument is candid and every citation says what the remarks claim.
- • Whether to interview the examiner, appeal or file a continuation instead.
ClaimMark-P is built so these checks are quick: every argument links to the passage it relies on, and where the model's interpretation is less certain, the draft says so.