1. How a requirement is marked
The AI marks each requirement in scope on its own. Each gets a score from 0 to 5 and a status of Met, Partly met or Not met.
2. Reasons and evidence
Each judgement comes with a written reason and the evidence it relied on, which is the document and, where there is one, the page.
3. Follow-up questions
Where evidence falls short, SmartAssessor drafts follow-up questions for the client. The assessor can add their own before the questions are sent.
4. Assessor review and sign-off
The assessor can override any judgement, and the override records who changed it and when. Findings are signed off by the assessor.
5. The report
Judgements, reasons and findings are compiled into an assessment report that can be exported as a PDF.
6. Data and models
SmartAssessor uses large language models from Anthropic, OpenAI and Google, and picks the model to suit the task.