How it works
How ALID discovers invention candidates from engineering work.
Connect once. ALID continuously reads what your engineers are already producing and delivers substantiated invention candidates to your review queue, with citations attached.
Why waiting for engineers to self-report fails
Engineers are not trained to recognize patentability. The standard intake process, a quarterly email asking engineers to fill out an Invention Disclosure Form, collects the inventions engineers already knew were notable, and misses every invention they didn’t recognize as one. ALID reads the engineering record directly so the pattern recognition happens on the attorney’s side, where it belongs.
Two ingestion modes
Automated integrations
Connect GitHub and Jira once. ALID continuously ingests new commits, pull requests, sprint tickets, and retrospectives. No engineer intervention. No tickets to file.
Google Drive and Confluence integrations are coming soon.
Manual upload
Drop in PDFs, Word files, PRDs, design docs, or meeting notes. ALID OCRs the content, extracts the technical claims, and runs the same pipeline. Useful for one-off discovery and for clients without standardized engineering tooling.
The 7-stage pipeline
- 1
Ingestion
GitHub commits, Jira tickets, and uploaded documents stream into ALID.
Attorneys see the engineering record directly, not a filtered version engineers chose to share. Nothing depends on engineers remembering to flag what they built or recognizing it as patentable in the first place.
- 2
OCR
Text is extracted from PDFs, Word files, and meeting notes.
Design docs, PRDs, and meeting notes that never make it into a structured ticket still feed the analysis. The corpus matches what your engineers actually produce — not just the slice that lives in a tracker.
- 3
Deduplication
Near-duplicate artifacts are clustered so attorneys never review the same idea twice.
When several engineers describe the same invention in different words across commits, tickets, and docs, ALID consolidates them into a single cluster. Your review queue surfaces inventions, not noise.
- 4
Grading
Each artifact is scored on technical novelty signals.
You see the highest-signal candidates first. Routine work — bug fixes, refactors, dependency upgrades, formatting changes — is filtered out before it ever reaches your queue, so review time goes to the work that might actually be patentable.
- 5
Mining
High-grade clusters are decomposed into discrete invention candidates.
A single complex feature often contains multiple distinct inventions. ALID separates them so each candidate can be evaluated, claimed, and filed on its own merits — rather than being lost inside a larger description.
- 6
Drafting
Each candidate gets a structured summary and a draft Invention Disclosure Form.
The first draft is written before you walk into the inventor meeting. The conversation shifts from "tell me what you did" to "confirm or correct what we found" — a far more efficient use of engineer time and a far more accurate record of the invention.
- 7
Packaging
Outputs land in your review queue with a confidence score and full source citations.
Every claim is grounded in a specific commit, ticket, or document — clickable and auditable. Nothing in your queue is unattributed, so attorney review can focus on legal judgment rather than re-establishing where each fact came from.
What you receive per candidate
Every invention candidate that lands in your queue carries the artifacts an attorney actually needs to act.
Structured invention summary
A 1–2 paragraph technical description in attorney-readable language, with the novel element explicitly called out.
Confidence score
A model-assigned confidence reflecting the strength of the underlying technical signal. You decide the threshold for review.
Source citations
Every claim is grounded in a specific commit, ticket, or document. Clickable, traceable, and reviewable in seconds.
Draft Invention Disclosure Form
A first-pass IDF, formatted for your standard intake. Edit, approve, or reject in your existing review workflow.
Time savings
Engineer engagement on a typical disclosure drops from an estimated ~20 hours of meetings, drafting, and revisions to ~2 hours of review on the AI’s output.
Estimated based on intake workflows observed during early rollout; not yet validated against production customer data.
Common objections, answered
ALID is designed to fit the disclosure workflows attorneys already run, without asking engineers to adopt another intake tool.
See ALID on your own engineering data
Connect a single repo or upload one PRD. We’ll show you the candidates we surface, with citations, in under 24 hours.