The Automation Stack I Actually Use in 2026 (And What I Cut)

The Tool Graveyard Behind Every Clean Stack

I have a folder on my desktop called "deactivated." Inside are twelve browser bookmarks to tools I once called essential. Zapier workflows I built and then unwired. An Airtable base I spent three days architecting that I haven't opened since October.

Nobody talks about this part of building an automation stack. Everyone shares the polished version — the well-oiled system where everything connects and nothing requires attention. The real version includes a lot of dead ends.

Here's what actually stayed, and why.

The Core Problem I Was Trying to Solve

Running multiple SaaS products — AutoChat, Hostao, RatingE, and a handful of smaller projects — with a small team means the operations overhead will kill you if you let it. Every manual task that happens more than twice a week is a tax on thinking time.

The specific pain points I was automating against: - New customer onboarding sequences that required someone to actually do something - Lead follow-up falling through the cracks on the hosting side - Content getting written but not published across multiple sites - Support tickets needing triage before the right person saw them - Weekly status visibility across products (where is everything?)

The trap I fell into early: automating things that weren't actually the bottleneck. Building clever zaps for processes that didn't need to exist at all.

What's Actually in the Stack

**n8n (self-hosted) for the backbone.** I switched from Zapier about 18 months ago. The honest reason: cost. At the scale I was running automations, Zapier's task-based pricing was adding up to something I could not justify against the value. n8n self-hosted on a small VPS costs me effectively nothing beyond the server. The visual workflow editor is comparable to Zapier. The main tradeoff is that complex integrations require more setup time — but complex integrations are exactly the ones where n8n's flexibility matters.

What I run through n8n: new lead routing from website forms to the right team member, customer signup triggers to onboarding email sequences, social media post scheduling from a content queue, and a few team status dashboards that pull from Supabase and render as Telegram messages to my phone.

**WhatsApp Business API through AutoChat for customer communication.** This is obvious given that AutoChat is our product, but it's also genuinely the right tool for our market. Indian SMBs are on WhatsApp. Reaching them on WhatsApp converts. I use it for Hostao renewal reminders, AutoChat onboarding check-ins, and RatingE client reports. The broadcast sequences have meaningfully better open rates than email for our audience — though I'd hesitate to give you a specific number because it varies a lot by message type and timing.

**Claude for drafting, editing, and classification.** I stopped using it as a replacement for thinking and started using it as a way to move from rough thoughts to usable drafts faster. The actual use cases that stuck: turning bullet-point notes from a call into a proper email, reviewing support tickets and suggesting response categories, and helping me draft RatingE report summaries when the raw data is already assembled. The use cases that didn't work: generating strategy documents autonomously, drafting anything that goes to customers without significant human editing, and anything where the input quality was low.

**Supabase as the data layer.** For anything where I need structured data across tools, Supabase beats the alternative of building fragmented state across different SaaS databases. Customer data, automation state, content queues, feature request tracking — it all lives there. n8n reads from it. The website UIs read from it. When something breaks, there's one place to look for the source of truth.

**Linear for task management, nothing else.** I tried every tool from Notion to Asana to plain text files. Linear is the one that's stuck. The keyboard shortcuts are fast, the automation around issue status is lightweight, and the team actually uses it. The reason it works where others didn't: it doesn't try to be a wiki, a docs tool, or a customer portal. It's a bug and task tracker and it does that well.

What I Cut and Why

**Zapier.** Cost-to-value ratio stopped working at scale. n8n covers 95% of the same use cases with a one-time setup investment.

**Notion as an operations hub.** Too flexible. When everything can be anything, nothing is optimized for a specific job. Great for documentation; not great for the kind of structured operational workflows I needed.

**A separate CRM.** We built a lightweight CRM view on top of Supabase instead of buying a third-party CRM. The savings were real, and the integration with our own systems was cleaner. This is probably only the right call for a technical team comfortable with Supabase.

**Four different AI writing tools.** I was paying for three separate subscriptions that all produced similar output. Consolidated to one. The differences between mid-tier AI writing tools are smaller than their marketing suggests.

**An email newsletter tool with a five-figure-subscriber price tier.** We had the subscribers but not the engagement to justify premium tier pricing. Moved to a simpler stack and the metrics didn't change.

The Framework I Actually Use Now

Before adding any new automation tool or workflow, three questions:

**What breaks if this doesn't exist?** If the answer is "nothing immediately, but it would be nice," that's a soft no. The tools that earn their place are the ones that, when you turn them off, something actually falls down.

**Am I automating a real process or automating around a bad process?** Most automation urges are actually symptoms of a process that should be eliminated or simplified. I spent two days building a routing automation before realizing the routing was necessary only because our intake process was unclear. Fix the process first.

**Who owns this when it breaks?** An automation no one understands is a liability. Every automation in our stack has a person responsible for knowing how it works and fixing it when it doesn't. If a tool requires expertise we don't have internally, that's a hidden cost.

The One Thing That Made the Biggest Difference

None of the tools above mattered as much as a simple Monday morning review of what broke last week. Fifteen minutes, in writing, every week.

What automation failed? What manual task happened that shouldn't have? What customer complaint traces back to a gap in the workflow?

The review creates the list. The list drives the automation priorities. Without it, you're either building things nobody needed or missing things that are causing daily friction.

I'm still not where I want to be on the operations side — there are manual processes in Hostao's support queue and RatingE's client delivery that I'd rather not have. But the stack is stable enough that I can focus on them systematically rather than fighting fires.

For anyone building a similar setup: [AutoChat](https://autochat.in) is where I'd start on the customer communication side, especially if your market is on WhatsApp. And [SuperLaunch](https://superlaunch.in) has useful frameworks for thinking about operations as you scale from solo to small team.

[Related: How I think about SaaS pricing across multiple products](https://reji.pro/blog/saas-pricing-mistakes-founders-learn-hard-way-2026)

Want to implement this for your business?

I help business owners build digital systems that actually work. Let's talk about your specific situation.