Unclear exam fees delay graduation
A non-profit educational institution posts 22,000 exam fees a month. Inconsistent details create thousands of clarification cases.
- Education
- Accounting
Key metric
Reduction in exam fee clarification cases
3,3001,500cases
Monthly clarification cases on exam fees before and after introduction
At a glance
Starting point
A non-profit educational institution with around 450 employees serves about 18,000 participants a year in courses, certificate programs and final exams. Each month, the back office handles around 22,000 exam fee bookings, spread across first attempts, resits, late registrations and cancellations. The fees originate in the course management system, online portals and manual forms. Discounts, subsidies and individual payment plans are documented in free-text fields or emails. The finance team works with a central finance system, several course management tools and spreadsheets to check and post the bookings.
- 22,000
- Fee bookings/month
- 450
- Employees in total
The challenge
Around 6,600 of the 22,000 monthly exam fees involve special features such as funding subsidies, installment payments, reductions or late registrations. The underlying rules are spread across policies, funding notices, course announcements and emails. The back office has to check every anomaly manually, gather supporting documents and resolve questions with course administration or participants. This creates about 3,300 clarification cases a month, which drag on for several days. Payment deadlines are missed, dunning runs have to be stopped and certificates can only be issued after manual approval. The team regularly works through peak periods at month-end and before exam dates.
Typical symptoms in the company
Why this remained unsolved
The problem had been known for years but was considered an unavoidable consequence of complex funding conditions and individual exam paths. A classic IT project seemed too big, because rules change frequently and many special cases are hard to fit into rigid input forms. Previous improvements were limited to extra spreadsheets and checklists, which tended to increase the workload. Fully harmonizing the process would have meant deep changes to course offerings and funding logic, which was not feasible politically or organizationally. So the process remained fragmented across the finance system, course administration and email, and heavily dependent on individuals.
Our solution
The institution introduced an agentic workflow that centrally bundles all exam fee transactions and supports the back office teams in checking, clarifying and posting. Incoming fee bookings from course administration, online portals and the finance system are brought together in a single workspace. An intelligent assistant reads free text, justifications and attachments, maps them to the relevant fee rules and proposes a classification: unremarkable, needs review or clarification case. For clarification cases, the workflow creates structured tasks with suggested questions, deadlines and next steps. After approval, the system writes checked booking texts and notes back to the finance system.
Where generative AI comes in
Generative AI analyzes free text in applications, emails and notes, recognizes fee rules, funding conditions and special agreements and condenses them into structured decision support. It drafts booking texts, case descriptions and questions, so complex special cases can be handled consistently without hard-coding every rule.
- Central worklist that brings together fee bookings from course administration, online portals and the finance system and sorts them by urgency.
- Analysis module that evaluates free-text justifications, funding notices and emails and suggests the applicable fee rule and the status of the case.
- Agentic workflow that creates clarification cases, drafts questions, monitors deadlines and writes structured booking texts to the finance system.
How the new fee workflow runs
Bundle data
Fee bookings from course administration, portals and the finance system are imported daily, enriched and shown in a shared worklist.
Pre-sort cases
An assistant assesses each booking based on rules and free text, marks unremarkable transactions and highlights potential clarification cases with a rationale.
Manage clarification
For clarification cases, the workflow creates tasks with suggested questions, deadlines and booking texts, which the team reviews, completes and approves.
Post and document
After approval, the system writes checked booking texts and notes to the finance system and updates the status in course administration and the worklist.
Learn and refine
Team feedback flows into patterns and suggestions, and new fee rules can be added without a development project.
Measured: impact after six months of operation
Period: impact after six months of operation.
| Metric | Value | Basis |
|---|---|---|
| Reduction in exam fee clarification casesKey metric | around 55% fewer clarification cases per month | System measurementmeasured |
| Processing time per clarification case | from 35 to 18 minutes per case | Time tracking before/afterestimated |
| First-attempt quality of suggestionsQuality measure | only minor corrections in 82% of cases | Samplemeasured |
Cost-effectiveness
The solution depends on the quality of the underlying course and funding data. If special agreements are only made verbally or recorded in unofficial notes, the workflow cannot reliably detect them and manual rework remains necessary.
Integration and operations
Transferability
The agentic workflow for exam fees shows how complex, rule-driven financial processes with a lot of free text can be relieved without forcing every special case into rigid input forms. Wherever bookings come together from several systems, individual agreements apply and supporting evidence sits in emails or documents, a similar approach can structure clarification cases, draft suggestions and free teams from repetitive searching and writing.
A similar case in your organization? Let’s talk.
Briefly describe your process. We’ll get back to you with an initial assessment of whether the pattern can be transferred, and what it realistically takes.