AI Glossary · Letter R

Reasoning Effort.

A setting on newer AI models that controls how long the model thinks before it answers, trading speed and cost for quality on harder problems.

Also known as Thinking budget, reasoning level, thinking effort

What it is

A working definition of Reasoning Effort.

Reasoning models work through a problem internally before they respond. Reasoning effort is the dial on that process. Vendors expose it as levels such as low, medium, high, and extra high, or as a token budget for thinking. Higher effort means the model considers more angles, checks its own work, and catches more errors. It also means slower answers and a bigger bill, because thinking tokens are billed even though you never see most of them.

Some products now set effort automatically based on the request. Many tools still leave it to the user or the developer, which makes it one of the most practical settings a team can learn to use well.

Why ad agencies care

Why Reasoning Effort matters in agency work.

Most agency tasks need far less thinking than the default, and a few need far more.

Low effort is fine for volume work. Resizing copy for twenty placements or tagging a folder of assets does not benefit from deep reasoning. Turning effort down makes those jobs faster and cheaper with no visible loss.

High effort earns its cost on judgment calls. Reconciling conflicting performance data, stress-testing a strategy, or checking a claims-heavy script against legal guidelines are where extra thinking catches real mistakes.

It is a quality lever people forget exists. When a team complains that a model gave a shallow answer, the fix is sometimes not a better prompt or a different model, just a higher effort setting for that task.

In practice

What reasoning effort looks like inside a working ad agency.

A strategy director uses a reasoning model to pressure-test a positioning recommendation before a pitch. On the default setting it returns a tidy but generic list of risks. She reruns it on high effort with the same prompt and the full research deck attached. The response takes a few minutes longer, but it spots that two of the supporting data points come from surveys with different audience definitions, a gap the team had missed. Meanwhile the same agency sets its asset-tagging and resizing workflows to low effort and cuts their run time noticeably.

Build AI workflows that actually run through The Creative Cadence Workshop.

The automations and agents module of the workshop teaches you how to build AI workflows that compress the busywork without taking the craft out of the studio.