Context
An insurance and extended warranty company with a high volume of settlements coming from technical services. Each file arrived with multiple associated documents: work orders, invoices, supporting documentation, approvals, and attachments in various formats.
The documentation entered heterogeneously — scanned PDFs, images, files renamed without standards, and backups with different criteria depending on the provider. The operations team manually reviewed each file before approving, noting, or escalating the case.
The problem
Validation depended on multiple manual controls: cross-referencing amounts, verifying documentation, detecting inconsistencies, and applying rules that changed depending on the type of service.
But the main problem was not the volume. Much of the operational criteria had never been formalized. There were exceptions that depended on who reviewed the file, rules that lived in the experience of a few people, and different criteria according to the type of technical service. As the volume grew, scaling the process meant increasing operational effort and keeping critical knowledge informally distributed within the team.
What we did
We built a document validation solution with AI agents: the agents cross-reference work orders, invoices, and supporting documentation, interpret unstructured information, validate business criteria, and generate automatic observations. When they detect inconsistencies or ambiguous cases, they escalate the file with context and already drafted observations; simple cases proceed automatically.
Specialized agents to understand complete files. We did not automate isolated steps — the agents interpret the case end-to-end.
Unstructured documentation processing. Technical services were not going to modify how they send information; the solution adapts to the existing reality.
Exception-based review model. The operator stops reviewing file by file and moves to intervening only when the case requires human judgment.
Iterative operational documentation. During the project, we identified and formalized exceptions and rules that had never been written down.
What we discarded
An automation based solely on rigid rules and traditional workflows. It would have worked for simple scenarios, but the process depended too much on variable documentation and implicit knowledge distributed across the operations team.
We also discarded demanding structured formats from external technical services: solving the problem by forcing changes on third parties would have added more friction than value.
Why it worked
We understood entire files, not isolated steps.
Humans intervene only when they add judgment, not as routine reviewers.
During the project, we formalized operational criteria that lived in the team's heads.
The solution adapted to how third parties already worked, without asking for external changes.
Stack
Specialized AI agents
Unstructured document processing
Operational rules engine
Integration with existing systems
Result
Pre-analyzed files before reaching an operator
The team stopped reviewing simple files from scratch
Operators focused solely on exceptions and complex cases
Operational criteria formalized and documented during the project
Critical knowledge that stopped depending on individual people
Agents do not replace messy processes. They expose them — and the more decisions rely on implicit criteria or undocumented exceptions, the more visible the problem becomes.