These beta agents are free to use inside ChatGPT. They’re designed to demonstrate what “governed behaviour” looks like: clear scope, consistent structure, and refusal outside declared limits.
The point of these betas is to make one thing tangible: governed agents behave differently to typical prompt-built agents.
Give an incomplete prompt. Ask for a checklist. Push outside scope. Notice how the agent asks for missing inputs and avoids confident guessing.
These agents are designed for Australian contexts and workflows: statutory reporting and professional study.
An Australian statutory accounts review copilot for SME financial statements under AASB RDR. Designed to support preparers and reviewers with structured findings and workflow-ready artefacts.
A governed study assistant for Certificate IV in Accounting and Bookkeeping (FNS40222). Built for learning, planning, and integrity-safe practice — not assessment answers.
An exam-focused tutoring agent for CPA Australia candidates. Structured study planning, concept teaching, synthetic practice, and progress tracking — governed for exam integrity and safety.
These additional betas are UK-focused or globally applicable study assistants. They help demonstrate that governed behaviour holds across very different domains.
A UK statutory accounts preparation and review assistant under UK GAAP (FRS 105, FRS 102 1A, FRS 102). Designed for disclosure completeness and structured reviewer findings.
A purpose-built AAT study assistant (Levels 2–4). Plans, teaches, drills, and reviews progress using exam-safe, synthetic practice and structured outputs.
A CIMA Management Accounting study assistant. Syllabus-aligned plans, teaching, drills, mocks, and progress analysis — governed through defined workflows and output contracts.
A CFA study assistant (Levels I–III). Adaptive plans, topic help, drills, mocks, and analytics — governed for ethics, exam integrity, and non-advisory behaviour.
These betas are examples. If you have a workflow that needs clear scope, rules, and predictable behaviour, we can build a governed agent tailored to your business.
AgentVon agents do not execute real-world actions or act as domain authorities. They support human decision-making by surfacing assumptions, gaps, and options — then stopping where judgment begins.
