Startups chronically over-scope infrastructure and under-resource it — this script helps engineers and CTOs have the conversation that prevents 18 months of pain.
## Why This Exists
The pattern: early-stage founder reads a HN post, asks for CI/CD + infra-as-code + multi-environment + zero-downtime deploys + monitoring. Assigns it as a side task to a solo dev or junior. No dedicated time, no budget, no operational ownership. Six months later: fragile one-off scripts, no runbooks, one engineer who knows how it works. This script gives you the words to have the scoping conversation before you're in that situation.
## When to Use It
Before agreeing to build or own infrastructure at a company where it's not clearly a first-class function. Works for an incoming CTO, a senior engineer taking on 'the DevOps stuff,' or a founder scoping what to build in year one.
---
## THE CONVERSATION FRAMEWORK
### Step 1: Expose the real ask by naming the levels
Say this: *'Before we scope this, I want to make sure we're talking about the same thing. There are four levels of infrastructure maturity. I want to know which level we actually need right now, not which level sounds good.'*
| Level | What it means | Realistic cost (eng hours to build + maintain) |
|-------|--------------|-----------------------------------------------|
| **L1: It runs** | Manual deploys, one environment, no IaC | 20h to set up, 5h/month to maintain |
| **L2: It runs reliably** | CI/CD, staging env, basic alerting, deploy rollback | 80h to set up, 15h/month to maintain |
| **L3: It runs at scale** | IaC (Terraform), multi-env, secrets management, on-call rotation | 200h to set up, 40h/month to maintain |
| **L4: Platform engineering** | Internal developer platform, SRE practices, SLOs, chaos engineering | 500h+ to set up, dedicated headcount |
### Step 2: Ask the calibration question
Say this: *'What's the cost to the business of 4 hours of downtime right now?'*
- If the answer is 'we'd lose some MRR' — you want L2.
- If the answer is 'we'd be in breach of contract with enterprise clients' — you want L3.
- If the answer is 'we're pre-revenue, mostly building' — you want L1 and nothing else.
Most pre-Series A startups get honest when you ask this. They realize they want L4 but can only justify L1.
### Step 3: Scope to the level they can actually own
Say this: *'I can build L[X] for you. What I need from you is: [list what you actually need]. Without those, I'll build it but I can't own it — and unowned infrastructure becomes a liability faster than you think.'*
For L2, what you need:
- One engineer whose primary job includes maintaining CI/CD (not a side task)
- A decision on cloud provider that won't change for 12 months
- A monitoring budget ($200–500/month for a small team is realistic)
- Agreement that deploy frequency is a metric leadership cares about
For L3, additionally:
- IaC is the source of truth — no manual console changes, ever
- An on-call rotation with compensation
- 20% of sprint capacity reserved for platform/reliability work permanently
### Step 4: The line you hold
If they want L3 without the resourcing for it: *'I can build L3 infrastructure. But if nobody owns it, it will decay to L1 reliability within 6 months while looking like L3. That's the worst outcome — it's expensive, complex, and fragile. I'd rather build you solid L2 that one person can operate than L3 that nobody can.'*
This is not an exaggeration. The 1,000-line Terraform script that destroys test environments when someone new runs it — that's what L3 infrastructure looks like when it's under-owned.
## The One Thing to Document After This Conversation
Write one paragraph: 'We agreed to target infra level [X]. The owner is [name]. The definition of done is [specific list]. We will revisit this scoping at [date].' Send it in email. You will need it in 3 months.