$11,000 a month of dispatch labor, replaced by an AI that costs $1.78 a month to run. A regional fiber ISP's field techs were calling a 10-person back-office desk for every equipment check, troubleshoot, and work-order closure, then sitting idle in their vans 30–45 minutes after 15 minutes of actual work.
Across 150 techs doing five visits a day, that downtime was the real cost. Nobody was measuring it.
What BlueLabel built:
→ A voice assistant giving techs natural-language access to troubleshooting docs and equipment data in the field
→ A dispatch agent that triages orders, runs hardware health checks against the OSS/BSS systems, and closes work orders with no dispatcher in the loop
→ Re-architected off the Custom GPT tier onto OpenAI's Agents API — which is what dropped infra cost to $1.78/month at scale
The results:
• Dispatch calls cut 50%
• $11K /month in labor savings
• 30–45 min returned to every technician, every visit
• AI running cost: $1.78/month serving 150 techs
The lesson: the headline is the $1.78. The value is the idle time that stayed invisible until someone built the thing that removed it. Make the expensive, unmeasured work visible first; then the automation comes second.
I broke down the full story in last week's Proof of Work: Applied AI Case Studies.
Full case study in the Pluris AI Use Case Library.
Full case study here →
checkpluris.com/case-study/h…
Our entire AI Use Case Library →
checkpluris.com/ai-use-case-…