The AI Marketing-Ops Stack That Replaced 5 SaaS Tools
Built once, runs forever, costs $42/month instead of $1,400. Here's the exact stack and what each piece replaces.
I rebuilt our marketing-ops stack last quarter. Started at $1,400/month across five SaaS subscriptions. Ended at $42/month with a setup that does more. Three months in, here's the actual stack, the actual numbers, and the two things I should have left as SaaS.
What got replaced
| Previous tool | Cost | Replaced by | |:--------------|:-----|:------------| | Enrichment (Apollo Pro) | $99/mo | Apify scrapers + Claude for inference ($8/mo) | | Lead scoring (HubSpot ML add-on) | $250/mo | n8n + Claude API ($12/mo) | | Sequencing (Outreach) | $140/seat × 3 = $420/mo | n8n + Postmark ($15/mo) | | Competitor intel (Crayon) | $450/mo | n8n + Firecrawl + Claude ($5/mo) | | Performance dashboards (Whatagraph) | $180/mo | Metabase self-hosted + Claude summarizer ($2/mo) | | Total | $1,399/mo | $42/mo |
What stayed as SaaS
Two things were genuinely worth keeping:
- GA4 + Search Console — free, nothing to replace
- HubSpot core CRM — the lock-in is too deep, the alternatives genuinely worse, and the cost wasn't the issue
If you're not using HubSpot, the question is real. If you are, don't try to migrate the CRM as part of an AI-stack project. Migrate one tool at a time.
The stack diagram
n8n (orchestration)
├─ Apify scrapers (data in)
├─ Claude API (inference layer)
├─ Postmark (email send)
├─ Slack (notifications)
├─ Metabase (dashboards)
└─ Postgres (source of truth)
Everything self-hosted on a $5 Hetzner box. The whole stack lives in Docker compose so I can move it in an hour if I need to.
The honest trade-offs
Time investment: 60 hours to build, ~4 hours/month to maintain. That's real time. If your hourly value is high, the savings might not be worth it.
Single point of failure: the n8n server. If it goes down, six workflows break at once. SaaS tools spread that risk across vendors. I mitigated with daily DB backups and Uptime Kuma monitoring, but it's still a real concern.
Onboarding new team members is harder. No "click here to learn the tool" — you're handing them a custom system. The CLAUDE.md and an hour-long Loom walkthrough have helped, but it's not as polished as Outreach's onboarding.
When I'd recommend this
Team of 1-10 marketers, technically comfortable, where SaaS costs are a real budget pressure. If you have 50+ marketers and SaaS is a rounding error, stick with vendors and don't make ops a side project.
What I should have done sooner
Started with one replacement, not five at once. Spent six weeks where the whole stack was held together with duct tape because I tried to swap everything in the same month. Replace one tool, run for four weeks, then move to the next. The pain savings are worth the slower timeline.
Get new tools first
Fresh templates, tools, and automation recipes in your inbox each week. No noise.
More from the playbook
- Multi-channel
AI Lead Scoring: Prompts, Pipelines, and the Pitfalls Nobody Warns You About
LLMs are weirdly good at lead scoring. Until they're not. Here's how to set it up, and the three failure modes that catch every team off guard.
- Email
Personalize 10,000 Emails a Week With AI Without Sounding Like a Bot
There's a precise line between 'they did their research' and 'this is clearly automated.' Hold it on the right side and reply rates triple.
- Multi-channel
Build an AI Ad-Copy A/B Testing Pipeline in 30 Minutes
Claude generates variants. Promptfoo scores them. You ship the winner. End-to-end, no spreadsheets.