Book a demo

Review

Your rulebook, applied to every document.

Check contracts, messages, claims and filings against your own policies, and point to the rule each issue breaks. Reviewers start from a flagged list with reasons instead of a blank page.

What it does

Built for tasks like these.

How it works

From your examples to a model you own.

  1. 01

    Collect past reviews

    Documents your team reviewed, with the rule each issue fell under and the ones that passed.

  2. 02

    Train on your rulebook

    The model learns your policies as your reviewers apply them, exceptions included.

  3. 03

    Flag before a person reads

    Every document arrives with its issues and the rules they break. Reviewers confirm instead of searching.

Private by default

Your contracts never leave the building.

The documents that need review are the ones you can least afford to share: contracts, client messages, claims. Reviewing them on your own hardware keeps every one inside.

Frontier API
ContractsInternetTheir GPUs
  • Your customers' data leaves the building on every call
  • The model changes whenever they ship
  • You pay for a huge general model to do one narrow task
Micromodel
your machineContractsModel
  • The request never leaves the device
  • The model is a file you own; it changes when you retrain it
  • Small enough to run on hardware you already have

Trained, not prompted

Accurate on your task.

Your policies have exceptions and house interpretations that aren't written anywhere a general model has read. Trained on your reviewers' past decisions, the model applies the rules the way your team does, and cites the rule it used.

InputClause 9.2: "Either party may terminate for convenience with 10 days' notice."

General model, prompted

This clause allows termination for convenience. You may want legal to review the notice period.

Micromodel, trained on your examples

§3.1 Termination · minimum notice is 30 days
Suggest: "with 30 days' notice"

Illustrative example.

And it costs less

No per-call bill.

Review queues are long and every item is a call. Locally, the thousandth contract costs what the first did.

$120,000 annual savings

See it for yourself

Watch a micromodel run on your own GPU.

A guided demo, about a minute. It downloads once and runs entirely in your browser; nothing you type is sent anywhere.
Open the demo

Sort and judge

Related solutions.

Have a task like this?

We'll tell you if a micromodel can do it. Or book a demo, or email hello@micromodel.company.