Agentic AI Cuts UX Friction Analysis Time 10x

The challenge

A digital transformation analytics company was drowning in its own data. Its enterprise clients generated billions of raw clickstream events, and the user friction hidden inside that scale — the small breakdowns that slow people down inside enterprise software — was hard to see, let alone fix.

Manual review couldn’t keep up, and it introduced its own problem: every analyst read the same data differently, so findings were inconsistent and every novel issue took repeated manual effort just to quantify.

What we delivered

SquareShift built a hybrid deterministic-plus-LLM pipeline on BigQuery, Vertex AI, and Gemini, organizing billions of user events into a single scalable data layer. An agentic AI architecture paired pattern recognition with causal inference and automated validation, linking UI and system errors to retries, pinpointing labeling errors, and quantifying time lost against an ideal user journey.

The system also surfaced forced workarounds and context-switching between apps — friction that was happening constantly but had never been measured.

The payoff

Friction analysis that used to take manual review now runs 10x faster, and because the pipeline is deterministic where it counts, findings are repeatable rather than dependent on which analyst looked last.

Hidden friction patterns get surfaced and prioritized automatically, so the team spends its time fixing high-impact issues instead of hunting for them.