June 30, 2026

When someone new joins our company, the first thing we should do is onboarding. Same as i had when i joined companies like Intel or Amazon Web Services (AWS).

We explain who we are. How we talk. Who our customer is. What we do and what we don’t do. How we want to be represented. That’s basic. It’s logical. Nobody questions it. But when we started using AI seriously on our team, we didn’t do any of that.

AI showed up, and we just started asking it for things.

The “new hire”

At Power Inside Out, we started with Perplexity. Then we moved to ChatGPT. Then we added Claude.

Every switch was hard. Internally, we joked and called it the “new hire”, someone who showed up without knowing how we work, with habits from somewhere else, and who took time to understand how we do things here.

What we didn’t see at first was that every new AI arrived knowing nothing about us. It came trained on companies from every industry, every kind of style, and information from millions of people.

Generic by nature. And on top of that, “contaminated” with everything we asked it day to day: travel questions, personal topics, general searches. All of that lived alongside our work. It didn’t come with the Power Inside Out model. It came with the model of the entire world.

And to make it worse, when the platform updated something (without warning, without explanation) the behavior changed. An analysis that included certain topics last week now brought up different ones. Same prompt. Different result.

It was like hiring someone through an agency, and every now and then the agency sent a different person to cover the same job.

The moment it clicked

When we started bringing new people onto the team, something similar happened. Each person came in with their own experience, their own style, their own way of talking to clients. And we had to put in the work so they could learn how we do things at Power Inside Out.

That makes complete sense with a person. We do onboarding, and we invest time so they can work well inside the company. We teach them.

And with AI? Nothing. Zero. We just let it improvise.

That’s when we asked ourselves the obvious question:

If I onboard a person so they understand how to work here, why wouldn’t I do the same with AI?

The solution we found

The answer came from AI itself.

We needed to create context documents (what we now call a “master context”) with key information about Power Inside Out: who we are, how we talk, who our customer is, how we run workshops, what our brand voice sounds like.

Now, before asking AI to do anything important, we simply load that document as context into the chat. The result is different. More aligned. More us.

There’s a second problem

When someone new joins the team, they bring their own AI. And that AI doesn’t know Power Inside Out. It has its own way of writing, its own style, its own history of conversations.

So the problem doubles.

The new person is learning how Power Inside Out works. And at the same time, their AI is going to produce content in the style of its previous owner, not the company’s.

Two learning curves at the same time. The person’s. And their tool’s.

The master context solves both at once. The new person loads the document, their AI reads it, and from day one they can produce material aligned with how Power Inside Out talks, writes, and shows up.

If something changes, we update the document. Everyone relearns, people and tools.

This is what we call, at Power Inside Out, building systems, not heroes. A process that works with or without you in the room, that anyone can follow, and that scales without depending on someone remembering how things used to get done.

What about you?

Every time you use AI without giving it context, you’re working with an employee who doesn’t know who you are or how you talk.

And it shows. Not in everything. But where it matters — how you communicate, how you present your company, how you talk to a client — it shows.

AI isn’t magic. It’s only as good as the context you give it.

Are you onboarding your AI, or are you letting it improvise?

If you want to know how we do it at Power Inside Out, or how to build your own system, message me. I’d be glad to talk it through.