AI Email Routing System 🤖

I build AI automation and workflow systems for copied errands.

I built an intelligent email routing system that reads incoming customer emails, understand their intent, allows them to the correct department, sends Gmail and Slack notification and prevent duplicate processing.

This is the complete workflow a control or Gmail trigger accept only messages carrying the copy test label.

I have to add this after discovered that a lot of unresited emails were coming into the inbox.

So I use that as like a control.

Then we have the second node which is the edit field.

It normalizes the sender, the subject, the body, the message ID and the date before using AI.

The duplicate guards actually check the Gmail message ID against the processed email table if it already exists.

The workflow stops the synchronization of that data without furthering wasting our time.

And then new messages are classified with the Google Gemini and instructed output parser which is the other node returns the department confidence score and reason.

The result then enters a deterministic switch router covering the sales logistics, billing, customer support and office admin.

This makes the classification quite easy.

AI undoes the interpretation while deterministic workflow logic controls the action.

Each branch sends the appropriate Gmail and Slack notification.

Then the mark process node stores the completed messages.

So I'm trying to do the end, the end testing to show you the results of the executions.

starting from the first node that was executed for the sales, a sale email was sent.

As you can see the execution was successful starting from the Gmail trigger, the edit fields, duplicates.

God.

And then it was classified by the Google Germany to show that the content of the email has to do with sales.

And then it was successful.

And then everything went.

The message ID was recorded here and then it was stored in the data table here.

And the next one is logistics through the same process as well as this is the billing, and then this next one is the support.

And everything that is happening here is also happening here in the Slack channel.

As you can see in the Slack channel you see all the mails that were sent to the email which I will show in a bit.

They all appeared in the, they got delivered, they arrived in the Slack channel as well.

Then if you look at the Gmail you see that we have five dedicated channels for the Gmail for the sales.

There is a Gmail for that.

You can see the email arrived.

So this shows that this actually worked end to end and then the final activity was done.

Has to do with the duplicate guard.

I sent a particular message that was sent earlier and it stopped at the duplicate guard because the message ID of that particular message has already been recorded earlier at the data table in the Mac process node.

So therefore it did not go further than this.

The workflow was stopped which means it was a real customer, system is going to prevent wasting of time and resources on that kind of email.

Thank you.

The final system combines contextual AI classification with deterministic routing control, Gmail intake, email and slack delivery, handling, retry logic and duplicate prevention.

I am Akita de Adroja and I, am an AI automation specialist.

So if you are hiring for your automation work or business processes which involves manual repetitive workflows, I am, available to discuss how I can help your business become more scalable and better.