AI output you cannot defend is AI output you cannot use.
A lawyer signs an advice. An accountant signs a return. An engineer signs a design. When AI helps produce that work, the professional still carries the responsibility, and "the AI said so" is not a defence. Renly makes AI-assisted work defensible: every response carries proof of what was asked, what the AI did, what it relied on, and which human accepted responsibility.
The reliance problem
Professionals are already using AI for real work. The gap is not capability, it is defensibility: six months later, can you show a client, a court, a regulator, or your own quality partner how that piece of work was produced and why you were entitled to rely on it?
A chat transcript is not that proof. It shows words, not actions. It does not show which sources were consulted, which systems were touched, who approved the writes, which model and policy governed the turn, or whether the record itself has been altered since.
The Evidence Passport
Every substantive response in Renly carries an Evidence Passport: open it from the response itself, or export it as an official PDF document.
Proves the action
The question and the answer, both content-hashed. The sources consulted and artifacts produced. Every tool action, with who approved each write. Sealed into the organisation's tamper-evident evidence chain, so alteration is detectable.
Explains the work
A structured decision trace: the task, the plan, each step taken with its result and duration. An account of the work a reviewer can follow, not a raw model transcript.
Names the authority
Which agent, which model, which provider, which policy verdict, and the governing instructions pinned by hash. Attribution to named users on every action.
Professional review, on the record
Reliance is a human decision, so Renly records it as one. A professional reviews a response and records the judgment: approved for reliance, reviewed with qualifications, or rejected, with the scope of review and any reliance limits.
The review binds to the exact content reviewed by hash and is itself cryptographically signed. If the underlying answer ever differed from what was reviewed, the mismatch is visible. Accountability stays with the human; the record proves how it was exercised.
The organisation's side of the story
AI Use Register
Why each AI use is permitted: purpose, accountable owner, risk tier, controls, lifecycle gates. Renly discovers your agents, workflows, and connectors automatically, so unregistered use is visible instead of invisible.
Proportional assurance
Higher-risk uses carry stronger requirements, up to fail-closed execution the organisation switches on per use. Runtime trust is earned from actual behaviour and drift is flagged, so an agent's standing reflects its record.
Disclosure, generated
An AI use disclosure statement generated from the register: purpose, oversight, data categories, and consent position per use. A working draft your privacy officer adapts, aligned with where Australian privacy reform is heading.
Work you can sign.
Renly is the governed runtime for teams whose work has to stand up later. Evidence on every action, judgment on the record, and an organisation-level register that explains the whole programme.
Talk to us Read: AI Accountability