Loading page…
Buyers send suppliers long security questionnaires, often hundreds of rows. This agent answers them from the company’s own policies. It cites the policy behind every answer, and passes anything it cannot support to a person, saying what is missing.
Step by step
Upload a spreadsheet or Word document in whatever layout it arrived. It works out which column is the question and which wants the answer.
It looks at how similar questions were answered on earlier questionnaires, which saves time and keeps the wording consistent.
Each earlier answer is checked against the current policies, then reused, updated if the policy has changed, or rejected if it claimed too much.
Every answer quotes the policy it relies on. The quote is checked word for word against the policy, so it cannot cite something that is not there.
Questions the policies do not support are not guessed. They go to a review list, each with a note on what is missing.
Judgment
These are the moments a simple search tool gets wrong. The demo data is built to include them.
A policy that requires two-factor login for administrators does not answer “Is it required for all users?” It notices the difference.
Some questions are only half covered. It answers the half it can and says what the other half needs.
If a past questionnaire promised something no policy backs up, it flags it, because a buyer may already have been told it.
Who does the work
A large, capable model that reads the sheet, checks the policies and writes each answer.
A smaller, faster model that asks whether the policy really covers the whole question.
Every agent here is built the same way: a leading AI model that plans the work and a smaller one that checks it, a set of tools each is allowed to use, and rules on what it must prove before it answers. They run on leading models from Anthropic or OpenAI, inside this website.
Safeguards
An invented software company, Tallowmere Systems, with twelve security policies and gaps in them on purpose. Its earlier answers include some that are out of date and some that claimed too much.