Replace SaaS Guide
Audit your SaaS stack, identify AI-replaceable tools, and migrate to self-hosted or custom workflows — with a cost spreadsheet and rollback plan.
The case for replacing SaaS
Most teams pay for 15–40 SaaS products. A meaningful subset — note-taking, scheduling, basic CRM, content drafting, internal search — can be replaced with open-source tools, AI workflows, or a single custom app. The goal is not zero SaaS. The goal is paying only for tools that are genuinely hard to replicate.
Step 1 — SaaS audit (spreadsheet)
List every subscription. For each row:
Tool | Monthly cost | Seats | Core job | API? | Replace difficulty (1-5) | Risk if removed
Sort by annual cost × replace difficulty. High cost + low difficulty = first targets.
Common quick wins:
| SaaS category | Replace with |
|---|---|
| Meeting notes | Whisper + local LLM summary |
| Social scheduling | Buffer clone or cron + API |
| Simple CRM | Notion/Airtable → Postgres + UI |
| Helpdesk (basic) | Email + AI draft replies |
| Design mockups | Figma stays; AI for asset gen only |
| Knowledge base | Git-backed Markdown + search |
Keep payment processing, email deliverability, and auth on proven vendors unless you have ops capacity.
Step 2 — Replacement decision matrix
Score each tool 1–5:
| Factor | Question |
|---|---|
| Frequency | Daily driver or quarterly? |
| Integration depth | Does 5 other tools depend on it? |
| Compliance | SOC2, HIPAA, PCI in the loop? |
| Build cost | Hours to replicate core workflow? |
| Exit cost | Export available? Lock-in clauses? |
Replace if: high frequency + low integration + good export + build cost < 6 months of fees.
Keep if: compliance-critical, deeply integrated, or replacement saves less than $200/month.
Step 3 — The replacement stack pattern
Most replacements follow this shape:
Data layer → Postgres / SQLite / Markdown in git
Automation → n8n, Windmill, or cron scripts
AI layer → Claude API, local Ollama, or OpenRouter
Interface → Retool, Streamlit, or Next.js app
Auth → Clerk, Auth.js, or SSO you already have
Do not rebuild auth, billing, and UI for a tool five people use. A spreadsheet plus AI plus Slack often wins.
Step 4 — Migration playbook (per tool)
- Export everything — JSON, CSV, Markdown, attachments
- Build parity MVP — only the 3 features your team actually uses
- Run parallel for 30 days — old SaaS stays active
- Cut over one team — smallest group first
- Cancel or downgrade — after zero rollbacks for 14 days
Never big-bang migrate production data on a Friday.
AI-specific replacements
| Job | Workflow |
|---|---|
| Internal FAQ | Embed docs in vector store; Slack bot answers |
| Report generation | Scheduled prompt → Google Sheet or PDF |
| Code review | Two-model review (see that guide) |
| Content drafts | Brief template → AI → human edit |
| Lead enrichment | Scrape + LLM summarize → CRM field |
AI replaces labor inside the workflow, not always the whole product.
Cost comparison template
Current: Tool X @ $49/mo × 8 seats = $392/mo = $4,704/yr
Replacement: VPS $20/mo + API $30/mo + 8 hrs build = $600/yr + one-time labor
Break-even: month 2 (if build ≤ 8 hrs at your rate)
Add 20% buffer for maintenance. If break-even exceeds 12 months, keep the SaaS.
Rollback plan (required)
Before canceling any subscription:
- Full data export stored offline
- SaaS account downgrade path confirmed (not delete)
- Team knows which URL to use if rollback needed
- Rollback tested once (restore one record, one workflow)
90-day roadmap
| Month | Action |
|---|---|
| 1 | Audit + pick 2 targets under $100/mo each |
| 2 | Build MVPs; run parallel 30 days |
| 3 | Cut over; log savings in one-page summary |
Keep payment, email deliverability, and auth on proven vendors. Never big-bang migrate on a Friday.