Why Shift-Left often just shifts the load

In practice, Shift-Left often does something else entirely: it shifts the cognitive burden onto people who are already overloaded.

Connections

🏅
Thank you for being a paying member.

This article follows "FinOps should be invisible"

Shift-Left sounds great on slides.

Push decisions closer to the teams. Empower developers. Move faster.

In practice, Shift-Left often does something else entirely: it shifts the cognitive burden onto people who are already overloaded.

If you truly want developers to make economically sound decisions, you must give them an enormous amount of context: pricing models, discount structures, commitment strategies, regional variance, internal standards, future scale assumptions.

That is not empowerment.
That is a second job.

Developers optimise for things that matter: correctness, reliability, performance, speed. Asking them to also optimise for cost does not remove work from the system. It adds work to their heads.

Yes, a developer may have a perfectly valid technical reason to choose one instance type over another. That choice should be respected. But it should exist inside a frame that already encodes company decisions.

That frame belongs one level up.

This is where architects matter.

Architects coordinate across teams and time horizons. They translate strategy into standards. They decide which degrees of freedom are open—and which are closed. Inside that space, teams can move fast without carrying the entire economic model with them.

Shift-Left assumes pushing decisions down automatically makes things better. Often, it does the opposite:

  • more decision fatigue
  • more inconsistency
  • more anxiety
  • worse global outcomes

Local optimisation with partial context is a reliable way to create expensive systems.

Good systems do not ask everyone to decide everything.
They decide once, encode it, and make the right thing the easy thing.

From a FinOps perspective, the goal is not to turn every developer into a part-time cloud-economist. It is to make cost-aware behaviour emerge naturally from well-designed defaults.

Shift-Left without context is abdication.
Shift-Left with full context is overload.

The sweet spot is neither.

It is decisions at the right level, automated everywhere else—and engineers free to spend their brainpower on the work that actually needs it.