Understanding how Attorly analyses documents helps you interpret results and decide which findings need action.
The model you choose
Each document is analysed by one AI model: the one you choose in your settings, or your organisation’s default. Attorly offers models from Anthropic, OpenAI, Google and Mistral; if you leave the choice on Auto, Attorly picks the model for you.
The model you pick is the model that answers. Attorly does not blend its findings with another model’s, and it never stands a different model in for the one you picked: if that model cannot run, you are told so and can choose another. The one exception is a model your organisation does not allow: Attorly never runs it, and the organisation’s default or another model it allows answers instead.
What the analysis contains
The analysis identifies the type of document, breaks it into its clauses and gives each clause a risk level measured against the governing law and the kind of agreement it is. It extracts the key commercial and legal terms, writes a summary and recommends what to change. It also points out provisions that are missing and parts of the document that contradict each other.
Limitations
AI analysis is a strong complement to human review, but it is not a substitute. The model may not know the commercial context of a transaction, the relationship between the parties, or why a clause was chosen. Always read the findings in the context of the whole deal and your client’s instructions.