For the past two years, “prompt engineering” has been treated as the defining AI skill. There were endless guides on magic phrases, secret prompt structures, and elaborate templates that promised dramatically better results.
In the early days, they often made a meaningful difference. They were the differentiator. They’re still an important part of my framework around AI.
But the landscape and the models have changed.
Today’s frontier models are remarkably good at understanding intent. Give them a reasonable request, and they’ll often infer the structure, ask clarifying questions, or even build the framework themselves. A prompt that once required a page of careful instructions can now be written in a sentence or two with surprisingly similar results.
Prompt engineering still matters. Good communication will always matter. But it’s no longer where the biggest advantage lies.
The new advantage is context.
From Better Prompts to Better Systems
The organizations getting the most value from AI aren’t necessarily writing better prompts. They’re building better systems or ecosystems.
They’ve documented their business. They’ve organized institutional knowledge. They’ve defined their voice, customers, products, and decision-making frameworks. They’ve connected their AI to the information that actually matters … and only what matters.
In other words, they’ve spent months building an ecosystem instead of minutes writing a prompt.
A company that’s actually done this has a living record of who owns which decision, a memory of why past calls were made and what happened afterward, and a standing way to tell its AI “here’s what’s changed since you last looked.”
A Recipe For Slop
The absence of that information and context is why so much AI-generated content still feels generic, and you see those artifacts of AI-construction.
It’s not because the models aren’t capable. It’s because they’re operating without context — their defaults come from the sum of the internet’s knowledge, not your organization’s actual preferences.
When an AI knows nothing about your company, your customers, your history, your goals, or your standards, it fills in the gaps with averages. It sounds like everyone else because, statistically speaking, everyone else is all it knows.
That’s where the telltale AI signs come from: generic introductions, predictable transitions, vague conclusions, and writing that feels polished but somehow empty. The model isn’t being lazy. It’s doing exactly what it should with incomplete information.
Think about hiring a new employee.
You could hire the smartest person in the world. Still, if you sat them at a desk with no onboarding, no documentation, no understanding of your customers, no explanation of your culture, and no access to the institutional knowledge your team has built over the years, you wouldn’t expect exceptional work on day one. You’d expect educated guesses.
AI is the same.
A clever prompt might take five minutes to create.
Someone else can copy it in five seconds.
They can reverse engineer it, ask another AI to improve it, or find dozens of versions online. Prompts have become increasingly commoditized.
Context isn’t.
Context is months of documentation. It’s years of accumulated knowledge. It’s your operating procedures, meeting notes, customer conversations, product documentation, brand standards, strategy papers, and the thousands of small decisions that make your organization unique.
No two companies will build exactly the same context.
That’s why context has become a competitive moat.
Perhaps the word “moat” overclaims slightly. A moat is static — dig it once, it defends forever. What the piece actually describes is closer to a flywheel that decays if you stop turning it. Institutional knowledge rots the same way any documentation rots if nobody keeps it current.
Context has to be maintained, not just accumulated.
The Next Competitive Divide
The gap today isn’t simply between companies that use AI and those that don’t.
It’s between organizations that have methodically onboarded AI into their businesses — creating systems where intelligent agents understand the company almost like a new employee—and organizations that are still opening a chatbot and typing random questions into a blank text box.
Those companies are technically using the same technology.
They’re just not getting the same results.
Twenty years ago, the differentiator might have been whether your business had a website. Ten years ago, until recently, it was social media presence … then social authority and podcasts. Recently, it was whether you had AI at all. Increasingly, that won’t be enough. The companies that pull ahead will be the ones that invest in building an AI ecosystem: one where knowledge is captured, context is preserved, and intelligent agents are equipped with the same information your best employees rely on every day.
The next phase of AI won’t be won by whoever writes the cleverest prompt. It will be won by whoever builds the best-informed systems.
Start with the one thing your best person knows that’s never been written down.
Onwards!
