Effective 17 October 2026 · Last reviewed 15 October 2026
Our commitments
- A person makes the final call. Our AI agents draft, calculate, summarise and propose. They do not file, submit, lodge or approve anything on anyone's behalf. A named practitioner, officer or reviewer decides.
- Output is checked before it is used. Each product verifies agent output before it can act: citations in Legal-OS, rule-based funding totals in Election Campaign OS, and confidence gates in ICPMO.
- Every decision leaves a record. What an agent proposed, what was checked and who approved it is recorded for audit.
- No training on client data. Legal-OS client data is never used to train AI models. Election Campaign OS refuses training use of personal information by default.
- Data location is stated, product by product. We say where each product stores and processes data, and we disclose every exception.
- Claims are tied to evidence. We describe what is built and measured. Anything else is labelled roadmap. We do not claim zero errors or perfect accuracy.
- Neutrality in civic products. Election Campaign OS offers every party the same modules on the same terms. We analyse wards, not people, and never predict results.
Raising a concern
If you believe one of our products produced a harmful or wrong result, tell us at legal@innovationconsult.co.za. We will check the audit record and respond with facts.