Human in the loop: AI proposes, people decide.
Human in the loop is the most important principle for AI in marketing and sales. Here I explain what the term means, which levels exist, where people should approve and how to implement it without slowing everything down.
What does human in the loop mean?
Human in the loop (HITL) means a person reviews, changes or rejects the output of an artificial intelligence before it turns into an action. The AI prepares and proposes; the human decides and is accountable.
The term comes from computer science and machine learning, where people check training data and correct models. Today it is used mostly for AI in companies: who may approve what before an email goes out, a proposal is sent or a campaign starts?
Last updated: 10 November 2026
Human in, on and out of the loop
| Level | What the human does | Example | Fits for |
|---|---|---|---|
| Human in the loop | approves every relevant decision | a follow-up email is checked before sending | anything that goes out or carries risk |
| Human on the loop | monitors, steps in when needed | AI sorts incoming requests, a person checks samples | internal routine with low risk |
| Human out of the loop | no control | fully automatic replies | rarely sensible in B2B sales |
Where people should approve: a simple rule
For every use case I ask two things: does the output go outside the company, and what does a mistake cost?
| Stays internal | Goes outside | |
|---|---|---|
| Mistake costs little | runs through, spot checks | a person approves |
| Mistake costs a lot | a person approves | a person approves, four-eyes principle |
The rules are written down per use case, with a named person. “Someone will check it” is not enough.
Human in the loop and AHI
AHI stands for Artificial Intelligence + Human Intelligence and is the principle THE Orange Growth works by. In THE Orange Triangle this level is called Control: above Brand, Business and Contact sits a steering level where a person takes the important decisions. Human in the loop is the mechanism, Control the responsibility behind it.
Implement it in four steps
- Collect use cases. Where do you already use AI, officially or not?
- Assess. Internal or external, low or high cost of mistakes.
- Define approval. Who approves, how fast, with which tool? In writing.
- Review and learn. Monthly: where did approval prevent mistakes, where did it only slow things down?
About the author
Jan Eric Bollig, THE Orange Guy.
Briefly answered.
What does human in the loop mean?
Human in the loop means a person reviews, changes or rejects the output of an AI before it turns into an action. The AI proposes, the human decides.
What is the difference between human in the loop and human on the loop?
In the loop: a person approves every relevant decision. On the loop: the AI works on its own and a person monitors and steps in when needed. Out of the loop: no human control.
Does the EU AI Act require human in the loop?
For high-risk AI the EU AI Act requires effective human oversight. Most marketing and sales applications are not high-risk, but human approval is good practice and lowers risk considerably.
What is AHI?
AHI stands for Artificial Intelligence + Human Intelligence, the principle of THE Orange Growth: AI delivers speed and data, people deliver judgement and approval.
AI yes, but keep control?
Let us work out together where a person must approve in your marketing and sales, and where AI can simply run.
Phone +49 173 545 39 88 · THEOrangeGuy@theorangegrowth.com