AI Usage Policy

1. The role of AI in Waysa

AI is an operational assistant. It helps your team do less repetitive administrative work by summarising documents, extracting structured information, drafting reports and suggesting next steps.

AI is not a decision-maker. The outputs you see in the Service are aids for human professionals. Every decision that affects a claim, a case file, a customer or a third party remains the responsibility of the human user.

2. What our AI features do

• Document summarisation. Multi-page PDFs, scanned documents and other case files are summarised into structured fields so a handler can see the gist before opening the file.
• Key-fact extraction. Entities (people, organisations, references), dates and amounts are pulled out of documents with citations back to the source.
• Missing-information detection. The system surfaces what appears to be missing from a file given the file type — for example, no medical report on a personal-injury case.
• Risk and red-flag scoring. For insurance files, the system surfaces fraud-risk signals with explanations and citations.
• Suggested next steps. The system proposes administrative next actions (request a document, schedule a review, send a reminder). These are suggestions only.
• Report drafting. The system can draft case reports and summaries from the underlying file. Output is editable and reviewable before it is sent or signed.

3. What our AI features explicitly do not do

• They do not give legal advice. The legal-workspace AI is an administrative assistant. It does not advise on the merits of a case, the strategy to pursue, or the rights and obligations of any party.
• They do not determine liability. Liability, coverage and settlement decisions are the responsibility of the qualified human user.
• They do not regulate professional conduct. Users remain responsible for compliance with the rules of their profession (SRA, FCA, GIA, etc.).
• They do not invent facts. Extracted dates, amounts and entities must be supported by the source document. When something is missing, the system says so — it does not fill gaps.
• They are not infallible. Large language models can produce inaccurate, incomplete or misleading output. Users must review AI output before relying on it.

4. AI providers we use

• OpenAI. Our primary large-language-model provider for summarisation, extraction and report drafting. Requests are sent over TLS to OpenAI’s API in the United States.

From time to time we may evaluate or introduce additional providers (for example, Anthropic) to improve accuracy, latency or resilience. Any new subprocessor will be added to our Privacy Policy before it goes live.

5. Training and reuse of customer content

We do not use customer content — files, summaries, reports or anything else uploaded to or generated within the Service — to train any machine-learning model, ours or a third party’s. OpenAI does not train on API content by default, and we have not opted in.

Inputs and outputs may be transiently logged by our AI providers for abuse-monitoring purposes in accordance with their own data-handling policies. We minimise the information sent to the AI provider to what is necessary for the requested operation.

6. Citations and auditability

Where the AI extracts a fact from a document, the system records the source so a reviewer can verify it. AI-generated content is marked as such in the interface and in any report exported from the Service. Each AI run is recorded in the audit log with the user who initiated it, the file involved and the model and response identifier returned by the provider.

7. Human-in-the-loop expectations

The Service is designed on the assumption that a qualified human will review AI output before any decision is taken or any external document is sent. By using the Service, you agree to maintain that review step within your firm’s workflow. Where the AI suggestion is wrong, the user is responsible for correcting it.

8. Prohibited uses

You must not use Waysa’s AI features to:

• Generate legal advice for end customers without independent qualified review;
• Make solely automated decisions that produce legal or similarly significant effects on individuals;
• Produce content intended to deceive, defame, harass or discriminate;
• Circumvent regulatory, professional-conduct or compliance obligations applicable to your firm.

9. Reporting concerns

If you encounter an AI output that appears materially wrong, unsafe or harmful, please report it to waysasystems@gmail.com so we can investigate and, where appropriate, improve the system.

10. Changes to this policy

As AI capabilities and the regulatory environment evolve, we will update this policy and the underlying Service. We will revise the “Last updated” date above and, where the change is material, notify customers in advance.

11. Contact

Questions about how Waysa uses AI can be sent to waysasystems@gmail.com.