Why Your Next Automation Might Not Need AI

Everyone asks for AI right now. But when a client describes their problem, the first question I ask isn't "where does the AI go?" — it's "does this decision need judgment, or does it need a rule?" If it's a rule, adding AI just adds cost and unpredictability to something that should be boring and dependable.

The problem

A third-party CDL driver recruiting agency runs on a pay-per-hire model — no placement, no fee. Once a driver's file moved into a carrier's processing stage, an account manager had to keep chasing paperwork and driver responses across several carrier-specific spreadsheets. There was no way to tell how often a file simply went quiet until someone happened to notice. Every file that went cold was a lost placement fee.

The build

The fix was a two-layer watcher with zero AI. A lightweight script inside each spreadsheet stamps the exact moment a file's stage or notes actually change. Then a scheduled n8n workflow runs twice a day and asks one question per file: has this moved since the last check?

If nothing changed, the account manager gets flagged and emailed with a link straight to the exact row that needs attention. If the file is still quiet after longer, a second reviewer is looped in. Automation only ever detects and flags — every real conversation with a driver stays human-to-human.

The gotcha

The naive version of this build would count every quiet hour against a file — including nights and weekends. That floods people with false alarms on Monday morning and teaches them to ignore the alerts. The watcher only counts business hours, so a file that sat quietly over the weekend isn't treated like a file someone ignored for three days. Alert fatigue is what kills monitoring systems, not missing features.

The result

Account managers stopped manually scanning sheets every morning — about 15 minutes saved per person per day, five to eight hours a month across the team. More importantly: recovered placements. At $1,500–$3,000 per placement, catching one or two files a month that would otherwise go cold pays for the build many times over.

No AI anywhere in the loop. Just a timestamp, a schedule, and a well-placed email.