Anonymised illustration. Based on requirements work I did at imcure. The 30% reduction is measured and on my resume; every other figure here is a labelled assumption, and no company data is shown.
Key findings
- A manual KYC flow required customers to upload documents, operations to re-key the data, and a person to verify it, with a rework loop whenever a document was invalid.
- Redesigning it into an assisted digital workflow cut manual verification time by 30% at imcure (measured, from my resume).
- The requirements that enabled it were small and testable: pre-fill from uploaded documents, route only mismatches to a review queue, and audit-log every decision.
Context
Customers needed identity verification (KYC) before activation. The process was manual: documents were uploaded, re-keyed by operations, then checked by hand, and invalid documents sent the customer back to the start.
Approach
I mapped the as-is process, marked hand-offs and rework, designed a to-be flow, and wrote the requirements as an illustrative excerpt with example user stories (acceptance criteria shown for illustration).
Analysis
Mapping the as-is flow exposes two hand-offs and one rework loop: every invalid document sends the customer back to the start, and an operator re-types data the customer has already entered. The to-be flow removes the re-keying and sends only mismatches to people.
- Customer uploads documents
- Operations re-keys the dataHand-off 1
- Manual verificationHand-off 2
- Valid?If no: customer starts again
- Account activatedAbout 3 days (assumed)
- Customer uploads ID once
- OCR extracts and pre-fillsAutomated
- Rule checks run automaticallyAutomated
- Pass?If no: exception queue with reason code
- Account activatedSame day (assumed target)
Source: anonymised process illustration based on work at imcure; times shown are assumptions. Analysis: Rajul Raman.
Recommendations
Pre-fill the form from uploaded documents, route only mismatches to a review queue with a reason code, and audit-log every decision. Illustrative requirements and user stories are in the appendix.
Impact
Measured: manual verification time down 30% at imcure (from my resume).
| KPI | As-is (assumed) | To-be target (assumed) |
|---|---|---|
| Time to activation | About 3 days | Same day |
| Manual touches per case | 3 | Under 1, exceptions only |
| Flow completion | Baseline | +15 to 20 points |
The assumed values illustrate how the redesign would be tracked; they are targets, not results.
Appendix
Illustrative requirements
| Ref. | Requirement |
|---|---|
| BR-3 | The system shall extract identity fields from uploaded documents and pre-fill the form. |
| BR-5 | Only mismatches shall be routed to a manual-review queue, each with a reason code. |
| BR-7 | All verification outcomes shall be audit-logged. |
Illustrative user stories
| # | Story | Acceptance criteria |
|---|---|---|
| 1 | As a customer, I upload my ID once so I do not retype details. | Given a clear ID image, fields pre-fill in under 5 seconds and stay editable. |
| 2 | As a customer, I see why a document was rejected. | Rejection shows a reason and a retry; no entered data is lost. |
| 3 | As an operations reviewer, I only see exceptions. | Queue holds only mismatches, each with a reason code. |
| 4 | As compliance, I need an audit trail. | Every decision stores who, what and when; exportable. |
| 5 | As a product owner, I track drop-off by step. | Dashboard shows completion rate per step, daily. |