1. Docs
  2. Build tests
  3. Analyze
  4. AI assessment

AI assessment

Assessment (AI) reads a main requirement together with all its sub-requirements and tells you how complete and testable they are. It scores the main and each sub, and lists what is missing.

3 min read Updated CustomersAdmins

In short

On this page you can

  • Score a main requirement and all its subs in one AI review.
  • See the areas no sub-requirement covers yet, as suggestions.
  • Read each sub’s own score, summary and suggestions.
  • Track progress with Last assessed and a fresh run after each edit.

How to use it

  1. Select the main requirement and save any changes.
  2. Open Assessment (AI) on the right and click Analyze.
  3. Read Quality Score, Summary and Suggestions.
  4. Turn each missing area into a new sub-requirement, save, and run it again.

Run an assessment#

  1. Select the main requirement

    Click it in the tree. Assessment works on main requirements only; a sub-requirement is assessed together with its main.

  2. Save first

    With unsaved changes, Analyze is greyed out: “Please save your changes before analyzing with AI”. Click Save.

  3. Open Assessment (AI)

    On the right panel, open the Assessment (AI) tab (on a narrow screen, the rail AI Assessment). The header shows “Main Requirement: name (ID: n)”.

  4. Click Analyze

    The button reads “Analyzing...” while the AI works.

    “Assessment completed successfully” and the result below. It is saved with the requirement and shown again next time.

Needs Enable AI in Project Panel › AI Options and a working AI key. A main requirement with no sub-requirements is assessed on its own.

What the AI checks#

One request goes to the project’s AI model with the main requirement and every sub-requirement. The AI answers two questions.

ScoredQuestions the AI is asked
The main requirementIs every aspect of the main covered by at least one sub? Is there enough detail across the subs? Are there gaps, missing flows or unhandled edge cases? Is the main itself clear?
Each sub-requirementClarity and specificity, testability, completeness, acceptance criteria, edge cases and error handling, security and performance, and fit with the main requirement.

What it receives: the names and descriptions of the main and its subs, and each sub’s severity. It does not see your App Map, Notes, documents or test cases, so it judges the text, not whether tests exist.

A judgement, not a formula

The score is the AI model’s verdict. Running it again can give a slightly different number even when nothing changed. Compare trends, not single points.

Read the result#

PartWhat it tells you
Quality ScoreThe AI’s score for the main requirement and how well its subs cover it, 0–100%, with a label: Excellent, Good, Fair or Needs Improvement. See how it is calculated.
Last assessedWhen it last ran.
SummaryOne or two sentences on the overall state. If the AI skipped some subs, it ends with “Note: n sub(s) could not be scored by the LLM.”
SuggestionsFor a main requirement: the missing areas, aspects no sub-requirement addresses yet. Each one is a candidate for a new sub.
Score BreakdownTwo bars and “n/m sub-requirements assessed”. If some subs were not scored, they are listed under “n sub-requirement(s) not yet assessed:”. See Score Breakdown.

Results for each sub-requirement#

The same run stores a score, a summary and 3 to 5 suggestions on every sub-requirement. Select a sub in Sub Requirements and open Assessment (AI) to read them. Analyze stays greyed out there: “Sub-requirements are analyzed together with the main requirement. Please select the main requirement and click Analyze.”

Raise the score#

  1. Add the missing areas

    Turn each suggestion of the main requirement into a sub-requirement with Add.

  2. Fix the weakest subs

    Open the subs with the lowest scores and apply their suggestions: a clear condition, a measurable result, the error case.

  3. Sharpen the main description

    State the goal, who does it, and what is out of scope.

  4. Save and analyze again

    Click Save, then Analyze.

Example · from Fair to Good
Before   Main "Pay with a saved card", 2 subs                Quality Score 48%  Fair
         Suggestions: expired card · card removed by the bank · 3-D Secure step
After    + 3 subs for those areas, sharper success criteria    Quality Score 76%  Good

Assessment and coverage are different things

A high Quality Score means the requirement is well written and complete. It does not mean it is tested. Tested is shown by linked test cases and Test Health; see What counts as covered.

If something goes wrong#

Didn’t find your answer?Write to support with your project name and what you see on screen.

info@smartestqa.ai