Home / Case studies / Forge

Case study · Forge in action

How KATO Forge delivered a 660% efficiency gain in email response

A multi-agent email automation deployment Kaze built inside a client's own Azure environment, showing what KATO Forge's multi-agent, human-in-the-loop capability adds on top of an existing inbox, CRM and knowledge base: from manual triage to machine-speed, policy-grounded replies.

Powered by KATO Forge

The challenge

Thousands of emails, one inbox, no shortcuts

Every customer email - billing, debt, metering, switching, benefit applications and more - arrived in the same inbox and had to be read, categorised and answered by hand before it reached the right team. The process stayed:

  • Reliant on an agent reading every email before it could even be routed.
  • Slow and costly at scale, with manual triage driving up first-response cost.
  • Inconsistent - reply quality depended on which agent picked it up.
  • Hard to scale without adding headcount as email volume grew.

What KATO AI adds

A multi-agent layer that reads, categorises and drafts a reply for every email in parallel - grounded in the client's own knowledge base and CRM, with a human reviewing every decision before it's sent.

Natural-language email triage ~600-article knowledge base CRM-grounded replies Full reasoning trail Human-in-the-loop review

How the multi-agent pipeline works

Every step is logged, explainable and reversible - hover each stage

1 · Ingest
→
2 · Enrich
→
3 · Score
→
4 · Recommend
→
5 · Human approve

First-response cost per email

Before vs. after KATO Forge - cost per reply is one part of a 660% overall efficiency gain

Before Forge £0.33 per email After Forge £0.05 per email 660% overall efficiency gain

Under the hood

One pipeline, every data source

Specialist agents read and reason across disparate systems in parallel - inbox, CRM, knowledge base and policy rules - rather than one brittle, point-to-point integration.

Email inbox CRM Knowledge base Policy rules Ingest Enrich Score Recommend Human review

Outcome

From manual triage to machine-speed, policy-grounded replies

660% overall efficiency gain

Lower first-response cost (~£0.33 to ~£0.05 per email) combined with parallel, multi-agent processing.

Lighter agent workload

Every email arrives pre-categorised with a drafted reply attached - not a blank inbox.

Scales without adding headcount

Multi-agent pipeline processes emails in parallel, not one at a time.

Back to the Forge capability overview