AI-native patent prosecution

Every rejection,
mapped to its source.

Prosecution Workbench turns a USPTO Office action into a structured, source-linked map — rejections, claim elements, cited art, specification support, amendment options, and estoppel risk. You move faster without losing the thread.

rejection claims elements cited art spec support amendments estoppel risk
Never files. Never guesses silently. Every assertion shows its source.
[0042] ¶ 7 · Fig. 3
§103 Claim 1 · element 2 arguably disclosed
Claim element
…a sensor configured to generate readings representing an environmental condition;
Examiner relies on — Smith, US 2018/0123456
“In one embodiment, the apparatus includes a temperature sensor 102 configured to generate temperature readings at regular intervals.”
Smith [0012] verified in source
Quote found verbatim · offsets 1184–1247 · nothing asserted that isn't there
The work behind the response

An Office action is an argument. Most of the work is making it legible.

Before

Twenty pages, three PDFs, and a clock

A final action arrives as a wall of prose, a stack of prior-art references, and a deadline. Which reference reads on which limitation, what the specification actually supports, where an amendment quietly costs you scope — that mapping happens in your head, or in a claim chart you rebuild from scratch every time.

After

The mapping, done in the open

Prosecution Workbench reads the action, lines the cited art up against your claims element by element, and shows the passage behind every assertion. The analysis is there to check, not to take on faith — and when the record doesn't support a claim, it says so instead of papering over it.

How it works

Structure first. The argument comes after.

The map is built before any prose. Draft language is rendered from it — never the other way around.

01

Parse the action

Rejections, statutory bases, the references cited against each claim, and the examiner's stated reasoning — pulled into structure.

02

Segment the claims

Each independent claim broken into its limitations. Edit the segmentation and the analysis downstream re-runs.

03

Map art to elements

Every limitation lined up against the cited art, each mapping quoting a real passage that is verified against the source.

04

Weigh the options

Amend or traverse, side by side — each path carrying its scope cost, new-matter risk, and estoppel exposure.

05

Draft the skeleton

Structured sections with USPTO-format amendments and a work-product appendix, exported to Word.

The claim chart & the strategy board

A source behind every line — and the tradeoffs in plain view.

The two views attorneys live in: what the art actually discloses, element by element, and what each response path costs. Sample data shown.

Claim elementCited art — verified passage
…a processor configured to receive the sensor readings;
expressly disclosed
“…a controller 110 receives the temperature readings from sensor 102…”
Smith [0013] · verified
…and transmit an alert when the reading exceeds a learned threshold.
missing / weak
No passage in Smith or Lee discloses a learned threshold; both teach a fixed limit.
gap flagged · the opening

The loudest treatment is reserved for what's missing — that's the attorney's opening.

Option AAmend — add “learned threshold”
Argument strength
Scope cost
New-matter risk
Estoppel exposure
Option BTraverse — no reasonable expectation of success
Argument strength
Scope cost
Support strength
Drafting burden

Options, not answers. Scope cost is weighed next to allowance — the call stays yours.

The line we won't cross

It accelerates the analysis. It never makes the judgment.

— does not file

Never files with the USPTO. The deliverable is an attorney-reviewable skeleton, not a submission.

— does not decide

No patentability, validity, or infringement calls. Those stay with the attorney, always.

— does not hide

Low-confidence and unproven analysis is marked, never quietly folded into a conclusion.

— does not surrender scope

Allowance isn't the only goal. Every amendment surfaces what scope it costs you.

Built for the doctrine

Fluent in the rejections you actually get.

§101 Subject-matter eligibility

Alice/Mayo step framing; features tied to a practical application or technical improvement, surfaced for review — never an eligibility verdict.

§102 / §103 Novelty & obviousness

Primary vs secondary references, the motivation-to-combine rationale, teaching-away and non-analogous-art angles, missing limitations.

§112 Definiteness & support

Written description and enablement, antecedent-basis issues, and §112(f) functional terms that may need structure in the spec.

Posture Non-final & final

After-final reality is surfaced — RCE, appeal, and continuation tradeoffs laid out for the attorney to weigh, not resolved for them.

For U.S. patent attorneys & agents

Read the four corners of the filing.
Keep the judgment beyond them.

Built for non-final and final Office actions across §101, §102, §103, and §112.