Real‑world chaos: Routing mixed sales/support emails from shared inboxes in SuiteCRM?

We have 3 shared inboxes (sales@, support@, billing@) connected via Inbound Email. Scheduler fetches ~200 emails/day.

Daily problems:

  • “Hi, interested in SuiteCRM customization quote” → ends up creating a Case in support queue (because no “support” in subject).

  • “Portal login broken after upgrade” → becomes a Lead assigned to sales (wrong filter match).

  • 30% emails create no record at all since they don’t match any workflow rule.

Reps spend 2+ hours daily manually fixing: moving records, reassigning, or creating from scratch.

Basic workflows (subject contains, from domain) just aren’t cutting it for free‑form customer emails.

Has anyone tackled this with smarter classification?

  • Custom logic analyzing email body to pick Lead/Case/Note?

  • Where to hook it? (Email::bean_importerSave(), InboundEmail scheduler override, after_import logic hook?)

  • External API integration examples?

  • How do you avoid duplicates or wrong assignments?

Production setups only please – what works at scale? SuiteCRM 7.12+ or 8.x.

Appreciate any battle‑tested advice! :folded_hands:

Hello Shah,

Meaning 1 inbox with aliases or 3 separate accounts / inboxes / credentials?
It should be 3 separate ones, then the scheduler can create cases for the support one.
Relying on people to follow rules / filter details won’t work - especially not when it’s internet traffic.
Best to split these up first, don’t rely on filters or similar logic rules.

You’ll still receive sales requests on the support email (and vice versa) from people who are just replying to your emails from the past.

If you absolutely have to have all 3 emails in one inbox (still: Don’t do that) disconnect it from the CRM and plug a middleware in between.
I’ve done quite some projects with PHP and n8n, the latter one being the faster approach.
Have n8n check every new emails, process it through NLP / AI (no cutting edge LLM required here), if you must, to classify and create / update / relate the records inside SuiteCRM via the API.

These are all CRM internal solutions. Better to tackle this problem before records are being created in the CRM.

  • External API integration examples?

Quite a few, yes:
https://www.youtube.com/playlist?list=PLsFGHEmMr2qf-H1COXZ2P2r_Bdg2BxYKY

  • How do you avoid duplicates or wrong assignments?

There are disadvantages to E-Mails via the API as well and you have to see how to handle those:

  • How to handle emails with two or more addressees?
  • Which message ID to store in the CRM (via the API, you’d not use the SuiteCRM generation logic for the ID)

We’ve hit this exact problem in production more than once.

SuiteCRM inbound emails are rule-based, not intent-based, so subject/from-domain workflows will always break with free-form emails.

At scale, tweaking workflows doesn’t help. You need to decide Lead vs Case vs Email before the bean is created, based on the email body, not just the subject.

What’s worked for us:

  • Intercept inbound emails during the InboundEmail scheduler (or a controlled before_save on Emails when parent_type is empty)

  • Analyze the full email body + sender context

  • Decide intent (sales / support / billing)

  • Then create the correct record and assign the right queue

This alone fixes wrong routing, dropped emails, and most duplicates (especially if you also check for existing open cases from the same sender).

Bottom line: workflows can’t understand intent — you need smarter classification earlier in the pipeline.

Hi @szee45, I see you are a Fynsis employee?

https://in.linkedin.com/in/shah-nawaz-ab2b0a19

Your post definitely looks like its cooked for SEO.

And now @fynsis comes in and drops the link to their own product (I edited the post to remove it), while writing things like

We’ve hit this exact problem in production more than once.

Why not just admit you’re trying to push your product?

And your product page about an Generative AI product has the following review:

Now this is truly brilliant. You have a new AI product, which is sold to companies, not end-users, and yet you already have 6k+ reviews!!! :laughing:

I will be watching your posts, or any posts, for that matter, and removing your SEO links, we don’t appreciate this way of doing things here. I usually just ban new users which do what you did - I only bothered to write this post for pedagogical purposes.

1 Like

I was surprised as well - maybe not an updated Linkedin profile or similar I thought.
But now with the reply as a solution, just a day later .. what a “coincidence”.