An anonymous company of the United States.
Automation of financial processes
Datawhale
Finance/business services
Explicit FDE engagement
Standard case | Useful reference with remaining gaps
Evidence level A
Deployed / scaled
A
Evidence level
Datawhale FDE Case 100 | No. 9: From spreadsheet-heavy work to intelligent public-enterprise finance execution
Evidence level measures whether a source can be located and reviewed; it does not mean vendor-reported claims were independently audited.
Standard case | Useful reference with remaining gaps
8 / 12
Business context1 / 2
Transformation workflow1 / 2
Technical workflow2 / 2
Human roles & governance1 / 2
Measured outcomes1 / 2
Source traceability2 / 2
Remaining gaps: Business context, Transformation workflow, Human roles & governance, Measured outcomes
Business problem
The processes of reimbursement, billing, audit, etc. are both semantic and subject to significant duplication of systems and high levels of accountability.
Solution
AI is responsible for information organization, classification judgement and risk identification; RPA/API is responsible for billing, filing and submission; and high-risk matters are upgraded manually.
Technical architecture & production workflow
Step 01
Operational materials/invoices/documents/applications
→
Step 02
OCR/ field extraction + structured business materials
→
Step 03
AI identifies the type of business, completes the context and makes the initial risk judgement
→
Step 04
Rule/supervisory logical review: amount, material integrity, conflict of rules, low confidence
→
Step 05
RPA/API entering ERP/OA to complete billing, fill-in form, attachment upload, submission
→
Step 06
High-risk/excessive review and final approval by finance staff
→
Step 07
Approval results, cause deposition as follow-up rules and examples
Key technology & infrastructure components
OCR/ Document ParsingLLMRules/monitoring logicRPA/APIERP/OA/Financial SystemHuman-in-the-loop
Human roles & accountability
Man-made responsibility for high risk, low confidence and final approval.
FDE delivery actions
- Watch the real financial operations with the guards and find out if the employees can't tell them what they're doing.
- Splits the process into three categories of "Semantic Judgement / Mechanical Execution / Final Responsibility"
- Interviewing old employees and translating experience into clear rules, exceptions and promotion conditions
- Reuse the original ERP/OA as much as possible and turn AI into an upper-level intelligence instead of re-engineering the financial system.
- Verify from PoC on small scenes and decide on private deployments and arithmetic inputs
- Design responsibility boundaries using low confidence and high risk routes
Reusable delivery patterns
- Enterprise AI is not equal to a PowerAgent; the most stable structure is often AI judgement + certainty enforcement + person responsible
- Renovation of a high-frequency process and expansion of infrastructure
- FDE's first product is often a new job stream, not a model.
Business outcomes & delivery results
The development of a transferable structure for “AI Sentencing + RPA Implementation + People Final Appeal”; some cases contribute to privatization numeracy inputs.
Evidence boundaries & verification notes
- : The Datawhale case discloses the division of labour and the method of landing; specific models, frameworks, interface products are not fully disclosed and components are indicated using generic technology categories.
- The source can be directly located in the case; the value remains disclosed by the source and does not represent an independent audit.