AI Usage Disclosure

• Sage (Sageon AI App) - the in-product assistant that helps you find prompts, explains workflows, and answers AI questions.
• Sage (Sageon Business) - the PMO advisor that reads your project data and can draft updates, log RAID items, move opportunities, and answer “what should I focus on this week?”
• SOFT report drafts in Sageon Business - AI-assisted first drafts of project SOFT reports (you approve before publishing).

All AI features run on Anthropic's Claude models, accessed via Anthropic's API. We do not use OpenAI, Google or other providers at this time.

Anthropic, PBC is a US company. Data sent to their API is processed in the US under the UK-US Data Bridge and Standard Contractual Clauses.

When you use a Sage feature, we send Anthropic:

• Your message to Sage
• Your conversation history within that chat session (up to 10 prior messages)
• A system prompt describing Sage's role
• A context block summarising your org's state - org name, plan, counts of projects/programmes/portfolios, RAG status, open RAID counts, pipeline values, recent project names, recent opportunities

We do not send:

• Salary or compensation data
• Employee personal data (full records)
• Full contact details (only counts are sent)
• Payment or billing information
• Authentication tokens or passwords

Anthropic does not use API data to train their models. This is contractually committed in their API terms.

API data is retained by Anthropic for up to 30 days for trust-and-safety purposes (detecting abuse) and then deleted, unless required to be retained for legal reasons.

You can read Anthropic's commercial terms at anthropic.com/legal/commercial-terms.

AI outputs are generated and can be wrong. Sage may produce inaccurate summaries, misremember project details, or make up numbers that look plausible but aren't in your data.

We mitigate this by:

• Grounding Sage in your real data via the context block (less room to hallucinate)
• Requiring confirmation before Sage writes any record (you approve every action)
• Showing what action Sage proposes before it executes

You should verify any AI output before relying on it for important decisions.

• You can choose not to use Sage features at all - it's opt-in by interaction
• Org admins can disable Sage org-wide via Settings → Modules (Sageon Business)
• Daily message limits (20/user/day) cap exposure automatically
• Any action Sage proposes requires your explicit confirmation

The EU AI Act came into force in 2024-2026, with most obligations taking effect by August 2026. Sageon's use of AI falls into the “limited risk” category - we believe we're compliant via this disclosure and our consent-by-interaction approach. We will update this page as the legal landscape evolves.

Questions about how we use AI: hello@sageon.co.uk

See also our Privacy Policy and Subprocessor List.