Pull the terms that matter out of physician employment agreements

A physician employment agreement buries its economics across thirty pages: base salary, the wRVU threshold and the rate above it, signing bonus clawbacks, tail coverage, the non-compete radius and its buyout. Reviewing one contract means finding each of those by hand; comparing three offers means a spreadsheet built from margin notes.

Thresh takes the agreement as a PDF and returns the terms you define as structured fields, each with the page it came from, so a reviewer can jump from any number straight to the clause behind it.

Ask in plain language, get conforming answers

Alongside fields and tables, you can ask questions the way a reviewer would: "Is the non-compete enforceable after termination without cause?" or "Who pays tail coverage if the physician resigns?" Answers come back grounded in the document, and anything the document does not answer says so explicitly instead of guessing.

The honest caveat about numbers: our published benchmark (98.6% field accuracy, 95.6% question accuracy) is measured on real estate and service paperwork, because that is where our first design partners live. It contains no employment agreements yet, which is exactly why the invite flow below asks you to send one of yours: you should judge us on your documents, not our averages.

How contract-review teams use it

• One template, every agreement. Define the terms once (compensation, schedule, restrictive covenants, benefits) and run every incoming agreement through the same template, so offers become comparable rows instead of thirty-page reads.
• A packet is one engagement. Group the base agreement with its amendments and exhibits into a deal and get one synthesized view, including discrepancies between what the amendment changes and what the base contract still says.
• Verification you can show a lawyer. Every value links to its source page with the clause highlighted. Review mode records who accepted or corrected each field, which is the audit trail a review practice actually needs.
• Straight into your system. Signed webhooks deliver completed extractions to your CRM or matter system. Our worked Salesforce example turns an uploaded agreement into native records, page numbers and confidence included, in about forty seconds.

On confidentiality

Compensation terms are sensitive. Documents stay inside our processing boundary, are never used to train models, and can be deleted on demand with an accounting tombstone left in place of the content. Retention windows are configurable per workspace, and an enterprise deployment can run inside your own environment.