Case study
Claims handling
Decision rationales are written by hand, case by case. AI drafts the text for the case worker and verifies every reference to the rules.
The problem
Companies that process complaints or insurance and damage claims must, for every case, write a decision rationale explaining why the claim was accepted, rejected or partially settled. To do so, the case worker has to manually look up and cite the relevant provisions of the contractual terms, internal guidelines or complaints policy, and write the text from scratch, case by case.
As the volume of cases grows, so does the risk of incorrect references to specific articles or sections of the terms. The case worker has to verify on their own that the cited provision actually exists and matches the case type — and write the entire rationale on top of that.
What this causes
- Hours of repetitive manual work writing rationales that follow a similar structure for similar cases.
- A risk of incorrect references to contractual terms or guidelines — a wrongly cited provision is both a complaint and a legal risk.
- Inconsistent quality of rationales depending on who handles the case.
- Slower case resolution, because the case worker first has to find the right provision before they can start writing.
What the client gains
- Shorter time to prepare a rationale thanks to an automatic draft based on the case data and the internal rule base.
- Zero risk of a non-existent or incorrect reference — the system verifies that the cited provision actually exists and applies to the given case type.
- A clear warning about non-compliance with the rules instead of an error slipping through unnoticed.
- From author to reviewer — the case worker verifies the text instead of writing it from scratch.
How we solve it
Connecting to the data source
We connect the solution to the internal system (CRM, case management system) that holds the case data and the base of rules and terms.
Validation and follow-up questions
The AI evaluates the completeness of the case data and actively asks for missing information (circumstances, case type, aggravating or mitigating factors) until it has everything it needs.
Generating the rationale
Based on the rule base and defined templates, the AI searches the relevant provisions and generates the text of the decision rationale.
Verifying references to the rules
The AI checks that every provision cited in the text actually exists and matches the case type. If not, it alerts the case worker and suggests a correction.
Review by an employee
The case worker checks the final text and copies it into the case handling system. Their role shifts from author to reviewer.
Further uses
- Automatic generation of rationales for other types of decisions (e.g. rejecting a warranty repair, returning goods, cancelling an order)
- Checking the validity of references to the terms and conditions in marketing and contractual materials as well
- Generating standardised responses for the customer line with a reference to specific provisions of the terms
- Automatic updates of rationale templates when contractual terms or internal guidelines change
- Extending decision rationales to other departments (HR, finance, legal)
Technologies
Python
Azure
SharePoint
OpenAI
Case studies
You might be interested
Znalostní asistent pro finanční oddělení
Odpovězení na dotazy z předpisů s citací a vždy z aktuální verze.
Invoice approval automation
Invoices that match the purchase order are approved automatically, with no manual intervention.
Automated label inspection
AI flags bad labels before they reach the customer.
