Automations & integrations

Turning a busy operational inbox into structured, assignable work, matched to the right property and issue.

Short
summary

Industry: Property management
Model: Back-office workflow automation
Stack: Gmail, n8n, Asana, language model
Users: Property managers, operations team

The team looks after dozens of buildings from one shared inbox. Residents, owners' committees and contractors all write to it: a lift that stalls, lights that are out, a reply asking whether anyone is coming. The automation reads every one of those emails and turns it into work in the right building's project.

The challenge

Someone had to read each email, work out which building and issue it belonged to, and copy it into the task system by hand.

An email is not a task. One message can report two problems, name no building at all, or be a reply to something already in progress.

The solution

A single n8n workflow sits between Gmail and Asana. It skips anything already processed, strips quotes and signatures, and asks a language model to classify the email and split it into one report per issue. A router then decides: a reply goes onto the task its thread started, a new report is matched to its building and filed, and anything ambiguous is handed to a person.

01Inbox

Mail keeps arriving where it always did.

Shared inbox with each email labelled by what the automation did with it

The same inbox on mobile, with processing labels

Visibility

3 labels

on every processed email: task created, follow-up logged or needs review

Once the workflow has handled a message, the inbox itself shows what happened to it, so nobody opens Asana to check.

An email reporting problems at two different buildings

The classification step with the cleaned email as input and structured reports as output

02Classification

The model reads the cleaned email and returns a fixed shape: the kind of message, a building hint, and a title, category and priority for each issue.

Structure

1 task

per issue, even when one email reports several

An email about a stalled lift and a dark corridor becomes two tasks with their own priority, so fixing the lift doesn't close the corridor.

03Matching

When the lookup is not sure, the model picks from a shortlist. An email that names more than one building is flagged for a person.

A review task for an email that names two buildings

The candidate-matching step choosing a building from a scored shortlist

Accuracy

3 steps

building code, then address, then name, before any AI is involved

Most emails match on a plain lookup. The model may answer that none of the candidates fit, so it never files a task under a building it invented.

A building's project in Asana with tasks created from email, by section

A task on mobile with its subtasks and a follow-up from the resident

04Tasks

Tasks arrive in the building's project with category and priority, ready to assign.

Context

1 project

per building, with every task carrying its source email

Each task has a summary and a set of subtasks, so whoever picks it up starts with the whole story.

05Follow-up

Replies are matched to their task by email thread and added as comments, next to the team's own updates.

A task with summary, original email, subtasks and a follow-up added by the automation

Continuity

Same thread

replies land on the task the first email created

When a resident writes back, the whole history of the issue stays in one place.

Outcome

The inbox stopped being the place where work was tracked. Every operational email now ends up as a task in the right building, or in front of a person when the automation cannot be sure.