The science

How PolicyAI turns policy wordings into answers you can trust.

No guessing from memory. PolicyAI reads the actual insurer wordings, finds the passages that matter, and answers only from them — then scores its confidence and cites every source.

The pipeline, end to end

Watch one question flow through the steps — each step opens up below as the flow reaches it. Click any step to explore at your own pace.

Step 1 · You ask
Compare home insurance across QBE, Allianz and CGU.

What is RAG, in plain English?

Retrieval-Augmented Generation means the AI reads the real documents before it answers — instead of guessing from memory.

✗ A plain chatbot

Guesses from memory

Answers from whatever it absorbed in training. It sounds confident, can’t point to your policy, and may invent limits or exclusions that aren’t in the wording.

✓ PolicyAI (RAG)

Reads the wording first

Retrieves the actual passages from the insurer’s PDS, answers only from them, cites the page, and tells you how confident it is — so you can verify every line.

Why you can trust it

Cited to the wording

Every claim links back to the insurer document and page it came from.

Confidence-scored

Each report shows how strongly the retrieved sources support the answer.

Honest “not assessed”

If no passage was found for something, it says so — it won’t fill the gap with a guess.

Grounded, not guessed

The model can only use the retrieved passages, which is what keeps hallucinations out.

Ask your own question

See a scored, sourced report for the cover you care about.

Open PolicyAI