Why we replaced every SaaS subscription with an AI agent

July 20, 2026 · 5 min read · philosophy

The SaaS trap

Every SaaS subscription is a recurring tax on your business. Email marketing: $99/mo. Customer support: $79/mo. Analytics: $49/mo. Hosting: $20/mo. Database: $30/mo. CI/CD: $50/mo. Scheduler: $25/mo.

Add it up. We're not talking about a few hundred dollars. We're talking about $400, $600, $1,000 a month — for software you don't own, can't modify, and could lose access to if the company folds or you miss a payment.

Enter the agent

An AI agent — one you can spin up on your own VPS, with your own credentials, and own the source code of — replaces a surprising amount of this. Not everything. But a lot.

We replaced:

Customer support ticketing ($79/mo) → a Telegram bot that answers FAQs, escalates to a human on demand
Email marketing ($99/mo) → a small script that batches sends via SMTP, with a small SQLite tracking table
Analytics ($49/mo) → Umami, self-hosted, no cookies, full ownership
Scheduling ($25/mo) → cron jobs, plain and simple
CRM ($129/mo) → Airtable, $20/mo flat — and even that we're replacing with a custom agent

The math

Before: roughly $1,000/mo in SaaS subscriptions, recurring.

After: roughly $30/mo in hosting (a single VPS), one-time setup costs amortized over 12 months.

That's $11,400/year saved. After taxes. Cash we can put into product, marketing, or just… keeping.

The tradeoff

It's not free. You have to:

Maintain the code yourself (or hire someone to)
Monitor uptime
Update dependencies
Handle edge cases the SaaS product "just handled"

But for builders — for people who already run their own VPS, who already write code, who already know how to debug — it's almost always worth it.

The rule of thumb

If a SaaS product is solving a problem you can describe in 200 words, you can probably replace it with an agent in 2 days. The exception is when the SaaS product's value is in its integrations (e.g., Salesforce syncs with everything) or its network (e.g., a marketplace).

For everything else — analytics, email, support, scheduling, basic CRM, basic marketing automation — the agent approach wins.

What we're not replacing

Stripe, Polar.sh, and payment processors. We don't want to build PCI compliance. We don't want to handle chargebacks. We pay Polar 4% of every transaction and we're happy with that trade.

The rule is: pay for things where regulatory exposure, fraud risk, or network effects make building yourself a losing game. Replace everything else.

— Digital Viking