A lawyer can build a legal decision tree without coding by defining one question, the facts that change its answer, and the outcomes each branch may produce. A visual builder turns those instructions into an assessment. The lawyer still supplies the interpretation, the exceptions and the points that require professional review. Start with a question you already answer repeatedly.
Take the instruction “send non-standard agreements to the right reviewer.” It sounds easy until two colleagues disagree about what counts as non-standard. The work starts with making that distinction explicit.
Legal decision trees are an established way to express rules. The AIJA and UNSW guide to automated decision-making describes systems in which experts define conditions and conclusions, including through no-code interfaces. That explains the mechanism; it does not establish that a particular legal interpretation is correct. AIJA and UNSW, AI Decision-Making and the Courts, sections 2.2–2.3.
Define the question and the stopping point
If you are still choosing what to package, first identify your first reusable legal assessment.
Choose a decision small enough to explain to another lawyer on one page. A candidate might identify which specialist should review a proposed transaction, which information is missing from an assessment, or whether a request falls within the scope of an existing service.
Write the intended user and the result together: “An associate answers these questions to identify the review route for a proposed agreement.” Then state what the result does not establish. In this example, selecting a review route does not approve the agreement or accept a new client.
Separate two kinds of input. A fact might be the proposed governing jurisdiction or the presence of an amended liability clause. A judgment might be whether that amendment is acceptable in the circumstances. If the judgment needs a lawyer, make that review part of the result instead of hiding it behind a yes/no question that the user cannot reliably answer.
You can start a schema in ProseID with that one decision. A schema holds the questions, conditions, sources and possible conclusions; a Flow is the experience someone completes. Keep the first version small enough that you can explain every branch.
Put the review boundary into the decision
Here is an illustrative internal policy for a law firm’s contract-review intake. These are invented practice rules for demonstrating the method, not rules of contract law:
- A request outside the practice group’s defined jurisdictional coverage receives Outside this review route.
- A request within coverage with a changed liability clause receives Specialist review required.
- A request within coverage where the user cannot establish whether that clause changed also receives Specialist review required.
- A request within coverage with no such change receives Standard review route.
“Standard review route” means the normal review can begin. It says nothing about whether the contract is enforceable or suitable for signature.
Give each question an answer the user can establish. For the clause comparison, that might be “Changed,” “Unchanged” or “Cannot establish.” Preserve an unanswered field as a separate incomplete state. Otherwise, the person who has not checked the clause can accidentally reach the same result as the person who checked and found no change.
State branch precedence as well. If a request is outside coverage and the clause changed, the outside-coverage route wins in this example. A different practice may choose a different rule. The author must decide that before testing the screen.
Build the result separately from the questionnaire
In ProseID, add the inputs as fields and the route as a calculated value. The visual schema builder guide explains how to describe conditions through the “If this, then that” interface and retain the calculated conclusion in completed Records.
Use field conditions to reveal the relevant questions. Use validation to identify missing or contradictory answers. Use a calculated value for the actual outcome. That separation matters: “Specialist review required” can be a valid, completed assessment, while an unanswered required question prevents completion.
For this example, a Determination is a sensible starting experience because the route is what the lawyer needs to obtain. The result should name the route, explain which facts selected it and give the next action. A Guided assessment may suit an interview whose main purpose is gathering information; the live demo lets you compare the experiences.
ProseID’s authoring tester allows you to inspect the calculation. The respondent receives the authoritative recorded outcome after a successful submission creates the Record. Do not put that outcome into an instructional message shown before completion.
Test the awkward cases before adding more questions
The table below is a proposed test specification for the illustrative policy. These are expected results to check when building it, not results from a deployed client tool.
| Test facts | Expected result | Reason for the test |
|---|---|---|
| Within coverage; clause unchanged | Standard review route | Establish the ordinary path. |
| Within coverage; clause changed | Specialist review required | Check the exception. |
| Within coverage; change cannot be established | Specialist review required | Give uncertainty a deliberate destination. |
| Outside coverage; clause changed | Outside this review route | Check which branch takes precedence. |
| Coverage unanswered | Completion blocked pending the required fact | Keep missing information separate from a substantive result. |
Ask a colleague to work through these cases from the written policy before showing them the software’s answers. A disagreement may expose an unclear instruction rather than a technical defect. Resolve the instruction, record the reason and repeat the affected cases.
Also test the wording. If a colleague must ask what “coverage” means, the field needs a definition. If they cannot establish whether a clause changed, the assessment needs a useful next step for that uncertainty.
Release a version another lawyer can understand
The assessment should keep the responsible author, its scope, supporting references and the interpretation used. In ProseID, a completed Record retains the submitted facts and calculated result against the exact published schema version. Later releases do not replace that historical version.
For an internal prototype, the first-Flow walkthrough explains testing and private publication. Before public release, review which questions, explanations and references can be disclosed. Use synthetic examples while designing the tool, and keep client-specific material out of public schema content.
Assign someone to maintain the interpretation. A new practice policy, source correction or confusing test case should prompt an explicit review. Version history preserves what the tool did; an accountable author decides whether its rules should change.
Finish with one question, a short outcome list and the test table. Once another lawyer can explain the result from those materials, the assessment has a basis for a useful first release.
This is a software-authoring method from ProseID’s founder. The contract-review policy is hypothetical. Legal review status: Not lawyer-reviewed. Product behaviour and linked sources checked 13 September 2026.