What Are Agentic Loops, & What Have They Got to Do with Marketing?
You may have seen a new AI phrase popping up: “loop engineering”. Just when we all got comfortable with prompt engineering. Let’s unpack it in plain language.
First, What’s An Agentic Loop?
Generally, AI is used sort of like a vending machine. You type in a request, you get an answer, and then you decide what to do next. You do all the steering.
An agentic loop works differently. An AI Agent can take actions, not just answer questions. In a loop, the Agent:
Looks at the situation
Decides what to do next
Takes an action, like researching or writing
Checks the result
Keeps going (or stops)
The cycle repeats until the goal is met or it hits a stopping point.
The agentic loop is the cycle. Loop engineering is the skill of designing it.
Think of a robot vacuum. You don’t steer it around every chair. It bumps into things, adjusts, keeps going, and returns to its dock when the job is done. (Without eating a phone charger, cat toys, etc., we hope).
So What’s Loop Engineering?
You often hear “agentic loops” and “loop engineering” used together. Here’s the difference:
the agentic loop is the cycle
loop engineering is the skill of designing it
The approach gained attention among software developers earlier this year (2026). Boris Cherny, who leads Claude Code at Anthropic, has described his job as writing loops that prompt Claude, rather than prompting it himself. Software engineer and writer Addy Osmani helped popularize the term “loop engineering” in a June essay, explaining the shift. (1)
Prompt engineering asks, “How do I write one good instruction?”. Loop engineering asks larger questions:
What should start the work?
What tools can the AI use?
How will we check the work?
When should it stop?
What needs the approval of a human being?
Loops need a clear finish line, and they can cost more.
What Does This Have to do with Marketing?
The term originated in the world of coding. But, the pattern fits any repeatable work with a clear finish line. And marketing has plenty of that! A well-designed loop could help a lean team with:
Monitoring: each morning, an agent checks brand mentions or competitor news and sends a short summary of only what changed
Feedback triage: an agent sorts customer feedback, groups it by theme and drafts replies for a human to review
Quality checks: an agent reviews a draft against your brand guidelines, fixes what’s off, and checks again until it passes
For a smaller team, that frees up time to work that humans do best: strategy, relationship building and big ideas. (2)
The Fine Print (and isn’t there always fine print?)
Are loops magic? No, but they are powerful.
They need a clear finish line. “Make our marketing better” isn’t a good prompt (in fact, this is an example we use at The Kendall Project, where I am a Certified Partner, but that’s another story. Learn more here.) A loop can’t check that prompt. Something along the lines of “Make every post have a call to action that is under 50 words” is.
They can cost more. Loops run the AI over and over. This uses more computing power than a single question.
Humans stay in charge (like we say at ChitChatDigital, “Don’t worry, the humans are still in charge…”. Anything that touches customers, your brand voice or your budget should get a human review first.
You still need to understand the work. A loop that nobody checks just makes mistakes more quickly. Not a good look.
Yikes! This Is All Moving Fast
Things are changing daily when it comes to AI technologies. Loop engineering is a new-ish idea, and the tools, terms and best practices move quickly. What’s true today may look wildly different next quarter.
The good news? You don’t need to build robots tomorrow. Start by identifying the repetitive, checkable tasks on your team’s plate. Those are your future loops.
And if you need help with marketing strategy, content, LinkedIn Executive Presence Management or AI training with The Kendall Project Framework, #LetsChitChat.
Authored by CMO Melissa Daley
SOURCES: