AI assistance with people accountable and in control.

What Ava assists with

• Capturing and qualifying enquiries across enabled channels.
• Preparing summaries, draft follow-up and next-best-action recommendations.
• Using approved tenant knowledge and workflow context to support relevant answers and handoffs.
• Showing confidence and source context where available so staff can review the basis of assistance.

Human control and accountability

AI is used to support staff, not remove accountability. Risky, external or high-impact actions should remain subject to authorised human review and approval.

Ava must not make final quotes, legal or compliance promises, finance approvals, availability guarantees or regulated decisions unless explicitly approved by authorised humans through an appropriate workflow.

Knowledge, confidence and correction

Tenant-approved knowledge and do-not-say constraints are intended to override generic AI responses. Confidence and source visibility help people identify where review is needed.

Human corrections can become reviewed improvement candidates, but they do not automatically become live tenant knowledge or training data.

Safety and training defaults

Logicl is building safeguards for prompt injection, unsafe requests, sensitive data and low-confidence responses. Suspicious or malicious testing should use authorised channels.

Customer conversations are not automatically used for model training by default. Any approved future reuse should be reviewed, appropriately redacted and tenant-scoped.