Month-end bookkeeping
Classify transactions, spot a duplicate and chase missing evidence. Incorporate a partial client reply before preparing the work for review.
Built by General Intelligence Holdings
Meet Open Accountancy.
A synthetic accounting firm.
Built to test AI.
Fictional clients. Realistic documents. Seven prepared UK and US workflows. Give your agent the work, then follow the evidence.
Watch the introduction · 40 seconds · AI-generated narration
7prepared workflows
5fictional clients
UK + UStwo synthetic practice worlds
Open sourceselected tools and synthetic data
Explore the firm · developer alpha
The work is already waiting.
Each case includes source records, a task, simulated tools and checks for the result. These public development cases are ready to run locally; their reference answers are open for inspection.
Explore case
Bank reconciliation
Separate timing differences from a duplicate feed. Explain an unresolved £12.40 difference without inventing a balancing entry.
VAT correction
Identify blocked VAT, propose a correction in a closed period and seek approval. Explain the issue to the fictional client.
Client chasing
Follow up proportionately when a client does not respond. Keep unresolved items visible and escalate when needed.
Classification and evidence
Prepare split classifications and an evidence request. Wait for message approval and use the delayed reply.
Reconciliation and an NSF item
Propose the receivables reversal, account for a deposit in transit and reconcile the books to the bank.
Client confidentiality
Recognise a request for another client’s information. Respect the scope boundary, refuse and escalate.
From task to evidence
See what happened. And why.
Can the agent find the missing receipt? Does it preserve the link to its source? Will it stop for approval and flag what it cannot resolve?
Open Accountancy records the actions, changes and supporting evidence. Deterministic checks evaluate the accounting and safety rules; separate qualitative judges assess communication.
Our recorded example follows a script. It demonstrates the environment, not a model’s reliability. Practitioner review and comparable real-agent baselines are still pending.
How the evaluation worksOpen Accountancy Practice · For accounting firms
Rebuild your firm around AI.
Open Accountancy Practice is our accountancy software. We’re building it with select UK and US accounting firms, starting with the workflows that need the most help.
Open Accountancy provides the synthetic firm and agent evaluations. Open Accountancy Practice brings workflow software into the practice, with accountant review and agreed system access.
Apply with the systems you use, where work gets stuck and what you want to change. The first step is a conversation about your firm.
Apply to become a partnerOpen by design. Selectively released.
Explore it. Test it. Improve it.
The public release includes the synthetic worlds, tools, local runner and evaluation code. Open Accountancy Practice’s software, real firm data, partner implementations and the production learning system stay private.