In short
- Context engineering is the work of deciding what your AI tools know, keeping it current, and making sure every team starts from the right version.
- It covers three questions: who manages the context, who manages access to it, and who decides how much each team uses.
- In bigger companies it looks like an operations team that serves everyone. In a small startup, one person has to own it.
- Until someone does, every team and every person grows their own context, and they drift apart.
I built the marketing context for my clients, packed it in a folder in their GitHub, and connected it to the AI tools their teams use. That's where the magic happens, and it's also where three big questions come up.
Answering them is context engineering.
What context engineering is
Context is what an AI tool knows about your company when it starts working: the positioning, the messaging, the product, the customers, the voice. Context engineering is the discipline of building that context, keeping it current, and getting it to the right tool at the right step of the work.
When the context is right, your AI tools write like your company, and you stop explaining the basics every time.
The three questions
Once a company works with AI across teams, these are the three questions:
- Who manages the context? Someone has to decide what goes in, what's true, and what changes when a decision changes.
- Who manages access to it? Not every team needs every file, and some of it is sensitive.
- Who decides how much each team uses? AI tools cost money per use, and someone has to set the limits.
There's also security, which deserves its own page.
It looks like operations
In larger companies, context engineering looks a lot like an operations team. It sits in the middle of the organization and serves everyone else: it sets up the context, manages changes other teams ask for, handles access and keeps an eye on usage.
The model that seems to be winning is hub and spoke: one central team owns the shared foundation, and every other team connects to it while working with its own tools and workflows. Marketing and engineering need very different context, and they still need the same facts about the company.
In an early-stage startup
A small team doesn't have a context team. Everyone does everything, especially around operations and AI. But someone still has to take responsibility, because until that happens, every team and every person grows their own context: one in the CEO's chat, another in the salesperson's, a third in a freelancer's tool. Each one drifts a little further from the others.
For marketing, the simplest version is one folder of plain files that holds the company's decisions, owned by the founder who owns go-to-market, and read by every tool the team uses.
Where to start
Start with the context that changes least and matters most: your positioning, your messaging, your tone of voice and your proof points. That's the core of your marketing brain, and it's what the positioning sprint leaves you with after 3 weeks, packed in markdown files with a few content-creation skills, ready for your AI tools.
How does your own homepage score? Get a free homepage audit: your homepage scored on the same 7 dimensions of positioning, with a line from your own page behind each score. Free, private, never published.