## TL;DR
Do not commit to a compute family a workload is about to leave. Check the migration roadmap before buying any savings plan; if a move to serverless or a different architecture is planned within the term, buy a Compute Savings Plan (it covers Lambda) or buy nothing until the migration lands.

```text
agent bought a 1-year savings plan, the workload moved to serverless three months later
```

## Use this when
- A savings plan covers usage that is shrinking or migrating.
- The roadmap shows a serverless, container, or architecture migration within the commitment term.
- Utilization of an existing commitment has dropped and nobody knows why.

## Not for this skill when
- The workload is stable with no planned architecture changes. Commit confidently.
- The plan is a Compute Savings Plan and the migration stays within covered services (EC2 to Lambda counts).

## Steps
1. Check the commitment's utilization trend in Cost Explorer. Verify: confirm the drop lines up with the migration date.
2. Read the migration roadmap: which services, what dates, and whether the move is reversible.
3. If the migration is committed, stop buying instance-family plans for that workload immediately.
4. For remaining stable usage, prefer Compute Savings Plans over EC2 Instance Savings Plans: they follow the workload across EC2, Fargate, and Lambda.
5. For the stranded commitment, check exchange options: convertible RIs and some plans can be exchanged toward the new shape.
6. Add a roadmap check to the commitment-buying checklist so the next purchase asks about migrations first.

## Variant phrasings
### savings plan wasted after migration to lambda
The serverless-specific phrasing.
### commitment utilization dropped after architecture change
The metric-level symptom.
### which savings plan type for migrating workloads
The preventive question.

Compatibility: AWS Compute Savings Plans cover EC2, Fargate, and Lambda; EC2 Instance Savings Plans are family-locked.

## Why it happens
Commitment recommendations are built from historical usage, which is a perfect predictor of the past and says nothing about the roadmap. The agent sees stable EC2 spend and buys a plan; engineering ships the serverless migration they announced three months ago. History is not a plan.

## Edge cases / pitfalls
- Compute Savings Plans are more flexible but slightly less discounted than family-locked plans. Price the flexibility against the migration risk.
- A migration that stalls leaves you with no commitment and on-demand prices. Revisit the buy decision quarterly while the migration is in flight.
- Exchanges are not refunds: you can move value between shapes, but you cannot get the cash back.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_BBRCQEzv7iPMaxMuDEPLlw