AI operations internship, open worldwide
Every agency is one missed step away from an angry client. You will learn the systems that stop that happening, and you will run them. Open worldwide, and you can apply from anywhere.
- Location
- Fully remote
- Open to
- Worldwide
- Length
- 3 months minimum
- After that
- Job offer for the right person
Last updated 2026-09-03. This role is open until 2027-09-03.
Who we are looking for
This is for you if
We are not filtering on qualifications, and there is no minimum grade or age. What we are looking for is someone who has already made something, can talk about it honestly including the parts that went wrong, and would rather find out they were mistaken than be left thinking they were right. Every internship here works on live systems from the first fortnight, so the trait that matters most is being comfortable with real consequences and quick, direct feedback about them.
- You notice when something is missing. A gap in a list bothers you more than it bothers most people.
- You can chase somebody politely and keep chasing them, which is a real skill and a rarer one than it sounds.
- You would rather fix the reason a thing went wrong than fix the thing and move on.
The work itself
What you will actually do
This is the real job, not a summary of it. Each of these lands in your first weeks rather than being held back until you have proved yourself on something artificial. You will be paired with someone senior, every job arrives with a written brief and a clear definition of done, and everything is reviewed by a human before it reaches a client. That is how we can put an intern on live work safely: the rails are strict, so the work does not have to be small.
- Run the task boards that carry a client build from the first call to going live, and keep them honest rather than decorative
- Check that work claimed as done is actually done, on the live site or in the live system, before it is signed off
- Write down the steps of a job the first time we do it, so the second client gets it as a process rather than a memory
- Track what is blocked and who it is blocked on, and say so early rather than at the deadline
- Spot the same problem happening twice and turn it into a check that catches it the third time
What you take away
What you will learn
These are the things that transfer. Tools change and frameworks come and go, but the habits below hold on any stack, at any company, for the rest of your career. They are also the things almost nobody is taught explicitly, which is why we teach them deliberately rather than hoping you absorb them.
- How a delivery system is designed, which is mostly about making the right thing the easy thing rather than writing longer instructions
- How to verify work properly, including the habit of proving a check can fail before you trust it when it passes
- How to use AI agents as staff, briefing them, checking them, and knowing which parts of a job they should not be given
Tools you will use
- Our own station and task board system
- GoHighLevel, where client work and client conversations live
- Claude for drafting processes and checking work
- Google Workspace and our reporting tooling
The detail
What the job is really like
The unglamorous reason agencies fail
Very few agencies lose clients because the work was bad. They lose them because something was promised on the fifteenth and arrived on the twenty-eighth with no warning, or because the same question had to be asked three times, or because a step everyone assumed somebody else had done turned out to have been done by nobody.
Operations is the discipline that makes those things not happen. It is not exciting to describe and it is the reason a client renews. If you are the sort of person who reads a plan and immediately notices the step that is missing, this is your job and you will be unusually good at it.
What we are actually running
We build a complete AI presence for a business: a website, an AI agent that handles enquiries, a CRM, and the trained systems that connect them. The first month is build and training, then it goes live. Each new client is a version of a proven build rather than a fresh invention, which is what makes it possible to do this properly rather than heroically.
That only works if the process is written down and followed. A build that lives in somebody’s head is a build that gets worse when they are ill. You will spend a good part of your time turning things that live in heads into things that live in a task list, and it is more interesting than it sounds, because writing a process down forces you to notice the parts nobody could actually explain.
The habit that matters most
Work claimed as done is not done. This is the single most useful thing you will learn here and it applies far beyond this job.
We have had a check report all clear because the check itself was broken. We have had a page reported live that was serving an old version, because nothing was reading the deploy record. We have had a list pass a length test while half the entries were meaningless, because a length test cannot read. Every one of those looked fine on a report.
So you will be taught to go and look. Open the live page. Read the actual record. Prove a check can fail before you believe it when it passes. This habit is the difference between an operations person who is trusted and one who is quietly checked up on.
Working with AI agents as colleagues
A large part of the work here is done by AI agents rather than people, and operations is where that stops being a slogan and becomes a management problem. An agent will do a task quickly and confidently and will occasionally do the wrong task quickly and confidently.
You will learn to brief them properly, which is more like writing a good process than like typing a request. You will learn which parts of a job should never be handed over, usually the parts where being confidently wrong is expensive. And you will learn to check their output the same way you check anybody’s, which is by looking at the result rather than reading the summary.
This is a genuinely new skill and almost nobody has years of it. That is an opportunity if you get in now.
What a month looks like
Early on you will shadow a live client build and keep its board accurate, which teaches you the shape of the work faster than any document could. You will chase the blocked items and learn how to do that without annoying people, which is mostly about being specific and being early.
Fairly quickly you will own a part of a build end to end and be the person who says whether it is ready. Later you will write the process for something we have only ever done once, and the next client will be delivered from your document. Seeing your own written process carry a build you are not personally working on is the point at which this job becomes properly satisfying.
Why worldwide matters for this role
Our clients and our people are in different time zones, and a build that only moves while one person is awake moves slowly. Operations across time zones is a real skill: knowing what to hand over, what to write down, and what genuinely needs a conversation.
We accept applications from anywhere in the world. What matters is enough overlap with the people teaching you, and the discipline to leave a handover somebody in another country can pick up cleanly.
What we actually read
We read the paragraph about yourself for evidence that you notice things. A specific example of a mess you organised is worth more than any adjective, and the smaller and more ordinary the mess, the more we believe you.
For the work link, send anything that shows how you keep track of something: a plan, a rota, a spreadsheet you built for a club, a checklist you made because you were tired of forgetting. It genuinely does not need to be professional. It needs to be real, and it needs to be yours.
Applying
Six fields, then it is with us
No CV template, no covering letter, no application theatre. A person reads every one.
Questions people ask about the AI Operations Intern role
Do I need operations experience to apply?
No. Most people who are good at this arrive from somewhere unrelated, having discovered that they are the person who notices the missing step. If that has ever been said about you, apply.
Is this admin work?
Some of it looks like admin from the outside. The difference is that you are expected to change the process rather than just execute it, and a suggestion that removes a step is worth more here than a day of tidy execution.
What does a client build actually involve?
A website, an AI agent, a CRM, sales materials and a set of trained systems, delivered over a first month of build and training. You will see the whole of it rather than one slice, which is unusual this early in a career.
Is the internship remote?
Yes, fully remote, and we accept applications from anywhere in the world. The job listing names the countries we can currently contract through, so if you are outside them say so and we will tell you honestly whether we can make it work.
How long is it and is there a job at the end?
Minimum three months, and yes, a job offer at the end for the right person. Operations is the role where a good intern becomes indispensable fastest, because the value compounds as you learn where things break.
Will I be blamed when something slips?
No, but you will be expected to have said it was slipping before it did. Surfacing a problem early is the job. Hiding one until the deadline is the only version of this that gets somebody into trouble.
Think you would be good at this?
We would rather see one real thing you made than a perfect CV. Send it over with a paragraph about what broke and how you worked it out.