The core bottleneck was manual search repetition
Paralegals were spending 15+ hours per week repeating the same search paths and still missing filings. The first fix was stateful checkpointing per court so the system asks only for new records instead of replaying old windows.
Pagination and dedup logic carried most of the workload
Filings often appear across overlapping result sets. We tracked compound record signatures to prevent duplicate inserts and stabilized output before delivery. Without this, the automation looked fast but polluted downstream sheets.
Delivery format mattered as much as extraction
The legal team needed daily summaries and per-court tabs they could trust. We pushed normalized rows and automated daily notifications so review started with decision-ready data, not cleanup work.