Start With the Audit
Before designing any automation, you need to understand what is actually in the inbox. This means categorising a representative sample of inbound emails by type: requests, notifications, approvals, inquiries, complaints, and noise.
In most organisations, this audit reveals that a significant proportion of inbound email falls into a relatively small number of categories. These categories are your automation candidates — the repetitive, predictable message types that consume time without requiring genuine human judgement.
Classification Architecture
Email classification can operate at multiple levels. The first level is intent detection: what does the sender want? The second is urgency assessment: how quickly does this need attention? The third is routing: who or what should handle it?
Modern language models handle intent detection well, particularly when fine-tuned on your organisation's specific communication patterns. Urgency assessment often requires additional signals — sender identity, subject line keywords, and historical response patterns.
Automation Tiers
Not all email automation is equal. Tier one is auto-routing: the email is classified and sent to the right person or queue without human triage. Tier two is auto-drafting: a response is generated and presented for human review before sending. Tier three is auto-response: for well-defined, low-risk categories, the system responds autonomously.
Most organisations should start with tier one and expand gradually. Each tier introduces additional risk and requires additional governance.
Measuring Impact
Track three metrics: the percentage of emails that are auto-routed without manual triage, the average time from receipt to first response, and the error rate (emails misclassified or misrouted). These metrics should be reviewed weekly during the first month and monthly thereafter.
Written by
The Orryx advisory team
Orryx is an advisory practice for AI and operational transformation. We work outcome-first and keep a human in the loop — our perspectives come from designing and governing automation in production, not from theory.