Automation · AI Classification

Inbox Triage That Labels, Alerts and Digests for a Crowded Mailbox

Every incoming email is classified against a fixed schema, labelled in Gmail and routed. Important mail raises an instant alert. Everything else waits for a scheduled daily summary, so the inbox stops interrupting.

Results at a Glance

3
Categories routed
1
Instant escalation path
1
Scheduled digest

The problem was not the volume, it was the interruption

Most inbox tools sort mail into folders and leave you to check them. That solves tidiness and not attention, because the checking still has to happen.

This one splits the two: anything important escalates immediately, and everything else is held and delivered as a single summary at a set time.

Industry Operations
Primary Result Instant escalation for important mail, one digest for the rest
Platforms
n8nOpenAIGmail
Services
Classification schemarouting rulesscheduled digest

The brief, and where the difficulty sat

A classifier that returns free text is unusable in a workflow, because the next node cannot branch on a sentence. The classification had to come back in a fixed shape every time, and the routing had to be explicit about what happens to each category.

Sorting does not stop interruption

Folders still have to be checked. Splitting urgent from everything else is what actually buys back attention.

Free-text answers cannot be routed

A workflow needs a fixed set of values to branch on, which means a schema and a structured parser rather than a prompt alone.

Misfiled important mail is expensive

The cost of a wrong classification is asymmetric, so important mail escalates rather than waiting for the digest.

A digest needs its own trigger

The summary runs on a schedule, independent of whether any mail arrived.

The challenge was real, and it was costing time, money, and client experience.

How it was built

A Gmail trigger passes each message to a classification step running on an OpenAI model with a structured output parser and a defined schema, so the result is always one of a known set of categories rather than prose.

A rules node then routes on that category: important mail is labelled and raises an instant alert, spam and promotional mail are labelled and left. A second scheduled branch collects the day’s mail, aggregates it and sends one summary.

The solution was designed from the ground up to fit this exact business.

Classify on arrival

A Gmail trigger passes each message to a model constrained by a classification schema.

Route by category

A rules node sends important, spam and promotional mail down separate paths.

Escalate the urgent

Important mail is labelled and raises an alert immediately.

Digest the rest

A scheduled branch fetches the day’s mail, aggregates it and sends one summary.

Tools Used

Each tool chosen specifically for reliability, integration depth, and ease of use for the client.

n8n

Classification, routing and the scheduled digest.

OpenAI

Classification against a fixed schema.

Gmail

Trigger, labelling and delivery.

What shipped

Urgent mail interrupts. Nothing else does.

1
Schema

A fixed output shape, so routing can branch reliably.

1
Alert

Important mail escalates the moment it lands.

1
Digest

Everything else arrives once, at a set time.

Before
Inbox checked continuously
Important mail found by luck
Promotional mail mixed with real work
No end-of-day picture
After
Every message classified on arrival
Important mail escalated immediately
Spam and promotions labelled and parked
One daily summary of everything else

Next Case Study

Video-Led Brand Site for a Prefab Villa Builder
Architecture · Prefab Construction
Video-Led Brand Site for a Prefab Villa Builder
A prefab villa maker that designs, manufactures, ships and reassembles in-house, on…

Need a build like this delivered for your agency?

Send me the brief, a Figma file, a reference site or a rough outline, and you'll have a fixed quote with a delivery date within 24 hours.