Open finance skill
Review past expense categories
Are past expense categories consistent enough to support automation?
Review expense categories by supplier, manual changes and ambiguous purchases. Check whether purchasing patterns changed over time before proposing an automatic rule.
Records you need
- ✓Categorized transactions
- ✓Consistent supplier names
- ✓History of manual category changes
What you get
- ✓Consistent candidates for automation
- ✓Conflicting past categories
- ✓Items needing accounting review
Checks performed
- ✓Enough history available
- ✓Consistent categories
- ✓Changes in purchasing patterns over time
From finding to decision
A manual category change may reveal new purchasing behavior rather than a broken rule.
Keep the original suggested category, approved result and reason. Without that history, it is harder to distinguish repeated mistakes from changed business needs.
- Compare current categories with prior suggestions and approvals for the same selected records.
- Group repeated exceptions by supplier, purchase purpose and available supporting records.
- Propose a specific rule improvement while leaving ambiguous or genuinely changed purchases for review.
Evidence ready for review
The history shows what changed, who accepted it and which records still need a decision.
Who makes the next decision?
Accounting decides whether to narrow a rule, provide better purchase records or continue manual review.
Collect the supporting records from your software
Level diagnostic method reviewed . Apply it to the selected records and accounting policy; the teaching example does not resolve a live account.
Guided synthetic example
See the check from input to decision
A vendor appears frequently, but its historical categories are inconsistent.
How review effort might be shared
An illustrative allocation is 15% collecting records, 20% applying checks, 10% flagging results, 25% investigating reasons, 20% finance review and 10% choosing an action. These percentages were not measured across engagements. Actual effort varies with data quality, complexity and the issue found.
Start with the source records
| Vendor | Transactions | Most common category | Consistency |
|---|---|---|---|
| Example Hardware | 62 | Materials | 61% |
| Sample Telecom | 38 | Telephone | 97% |
Apply the documented check
Measure normalized-vendor history, category concentration, overrides, and time drift before proposing automation.
Surface the flagged result
Needs review
Example Hardware is only 61% consistent and should not receive a broad automatic category rule.
Use recorded finance judgment to analyze possible reasons
At Level, we record the review logic, known explanations, and questions our finance professionals use in this situation. AI applies that documented human judgment to organize the most plausible reasons, without pretending it knows which reason is true.
Possible reasons to investigate
The vendor may legitimately span materials, tools, repairs, and office purchases, or the historical books may contain miscoding.
Make the final human judgment
In a Level engagement, a Level finance professional validates the source evidence, challenges the AI-assisted analysis, and decides which explanation is supported.
Make the operating decision
Keep the vendor manual or add a narrow secondary condition after finance policy review.
Each skill uses an independent synthetic scenario. Shared month labels do not mean the examples describe one company or a single financial period.
Use this system
Start with the example, then inspect the structure
The JSON download is a sample record file to inspect with an AI assistant or developer. It is not a file to upload into your accounting software.
- Download the invented input and expected output.
- Replace one field at a time with a safe test value.
- Compare the result with the expected structure.
- Have a finance professional review every exception before acting.
Website example JSON and registry: CC BY 4.0 (attribution required). Downloadable toolkit ZIP: MIT. Each version identifies its own frozen examples.
Developer preview: AI-assistant helper (MCP)
This tool call returns fixed, invented sample records only. It does not connect to your books. Read the setup instructions.
run_synthetic_example {"slug":"categorization-history-health"}Human review is part of the system
This check organizes evidence and surfaces exceptions. A person with finance domain expertise still needs to interpret the result, validate the source records, and discuss the operating decision.
Open source does not mean open client data. We publish reusable finance tools, checks, templates, and playbooks. Client records, workpapers, credentials, communications, and private configurations remain confidential.
Use it yourself, or bring us the messy part
Get a one-page plan for checking your records
Tell us where the records stop making sense. We will reply with the first check to run, the records it needs, and the decision that still needs human finance judgment.
Install the free skills yourself