Agentic AI saves Metcor 80+ admin hours a month and cuts ticket processing time by 93%.
A production agentic AI workflow, live inside Microsoft Teams and powered by Anthropic's Claude Haiku, that turned a 30-minute manual data-entry task into a sub-2-minute process — and keeps overhead flat as ticket volumes grow.
A manual bottleneck that grew with every van on the road.
Metcor is a leading UK field services provider specialising in HVAC, plumbing and refrigeration. As the business scaled, processing parking tickets for its field service vehicles had become a costly, manual bottleneck. Each ticket took up to 30 minutes of repetitive data entry, with staff transcribing details by hand from scanned documents.
The process introduced errors and pulled valuable staff time away from strategic work. With ticket volumes rising, and overhead costs at risk of growing in line with revenue, Metcor needed a scalable way to streamline operations without compromising accuracy or compliance.
- Manual processing took up to 30 minutes per ticket, hurting productivity and delaying Field Service Management (FSM) workflows.
- Scanned PDFs and images caused frequent data inconsistencies because there were no intelligent extraction tools in place.
- Scaling exposed the limits of manual workflows, and any fix had to work inside Microsoft Teams, where staff already worked every day.
A lightweight agentic workflow, built where the team already works.
We built an AI-powered solution inside Microsoft Teams using the Teams Toolkit. Staff upload scanned tickets straight into Teams, which triggers a multi-agent workflow orchestrated with LangGraph and powered by Claude Haiku, Anthropic's fastest model. The agents parse each document, extract the relevant data and produce structured output, with a person confirming the result before anything is submitted.
Upload in Teams
Field staff drop a scanned ticket or photo into a familiar Teams chat. No new app to learn.
Claude agents parse & extract
LangGraph orchestrates Claude Haiku agents that read the scanned document and pull out the fields that matter.
Human review
Users confirm or adjust the AI output before final submission, which keeps accuracy high and builds trust.
Excel-ready export
Validated data is exported in an Excel-compatible format, ready for existing FSM processes.
Logic-driven orchestration
LangGraph gave us scalable, controllable automation, so every ticket follows the same reliable path.
Human-in-the-loop by design
A review step keeps people in control of what gets submitted, which matters for accuracy and compliance.
Zero-friction adoption
Because it runs inside Microsoft Teams, it fitted into existing FSM processes without disrupting them.
Fast, accurate, and cheap enough to run at volume.
- Speed: keeps each ticket well under two minutes end to end, review included.
- Reads scans directly: no separate OCR step to maintain.
- Cost at scale: admin costs stay flat as ticket volume grows.
Faster, more accurate, and built to scale without extra headcount.
Business continuity maintained
Deployed inside Microsoft Teams, a tool already in daily use, the solution caused zero disruption to operations. No training or platform change was needed, so adoption was fast and returns were immediate.
Performance optimised
Each ticket is now handled in under two minutes, an improvement of over 93%. The system reduced errors, kept data formatting consistent and set a benchmark for similar administrative use cases.
Costs optimised
Manual processing led to missed deadlines, delayed recordkeeping and admin overhead. The automated workflow cut that overhead and showed that such processes can scale without a matching rise in headcount or admin costs.
What stood out wasn't just the time saved — it was proof that we can automate a repetitive process without adding risk or losing control of it. It's given us the confidence to look at agentic AI for other parts of the business.
Have a repetitive process AI could take off your team's plate?
Tell us the workflow that's slowing you down. We'll tell you whether agentic AI is the right tool and what it would take.