Our internal company surveys yield polite corporate generalities instead of real feedback. How can we use a scraper and language model on our internal support logs to uncover genuine operational friction points?

Question

Tyler Smith · Accepted Answer

Company surveys fail because employees self-censor when asked for feedback directly. They do not want to sound negative or blame coworkers, so leadership receives bland suggestions about office snacks while core operational handoffs remain broken.

To find the actual points of operational friction, look at where your team confesses frustration in real time. Deploy a simple scraping script using Playwright or an internal data export to pull text logs from your internal IT helpdesk, customer escalation channels, and warehouse dispatch chats over the past six months.

Run these unstructured logs through an LLM configured with a clear operational prompt. Instruct the model to ignore routine pleasantries and isolate recurring operational complaints. Specifically, direct the model to extract instances where team members report waiting on files, missing passwords, incomplete intake forms, broken software handoffs, or repeated manual rework.

Group the extracted issues by process and frequency. The analysis will immediately reveal hidden bottlenecks that never surfaced in quarterly reviews, such as customer handoffs that stall because sales omitted critical billing codes or field technicians waiting hours for updated dispatch sheets.

Bring this synthesized list of systemic friction points directly into your next quarterly planning session. Use the data to prioritize Rocks that streamline broken Core Processes. When you base operational decisions on unfiltered workflow logs rather than polite surveys, you fix the real friction holding your team back.