What is "context" and why does AI need it?
Most people who try ChatGPT for the first time get one of two reactions. Either it feels like magic, or it feels generic and disappointing. The thing that decides which one you get is context.
Context is just the background. Who you are, what you do, who you're talking to, what's already been said. It's everything an AI tool needs to know before it can give you an answer that's actually useful, instead of a polite-sounding average of the internet.
Why generic input gets you generic output
If you ask ChatGPT to "write a follow-up email to a customer," you'll get back a competent email that sounds like it was written by a marketing intern at a SaaS company. That's not ChatGPT being bad at writing. The tool can only work with what you gave it, and you gave it almost nothing.
Try the same request again, but this time add: the customer's name, what they bought, what went wrong, how you usually sign off, and one quirky thing your team always says. You'll get something that's 80% of the way to what you'd actually send.
Same model, same prompt structure. The only thing that changed is what you fed it.
What "good context" looks like for a small business
For most of the small businesses we work with, useful context falls into four buckets.
Who you are. Your company name, what you do, how long you've been doing it, your usual tone. We've seen this written as one paragraph, and we've seen it written as a 12-page brand voice doc. Both work, as long as the AI actually has it.
Who you talk to. Your customers, your team, your suppliers. What they care about. What they already know. A roofer doesn't need the same email as a software CTO, and an AI that doesn't know who's on the other end will default to "vaguely corporate human."
What's happened before. Past emails, past quotes, the last conversation. This is where most AI tools fall flat without help, because the chat window forgets everything the second you close it.
Why this matters now. What you're actually trying to do. Get paid? Reschedule? Apologize? Win the bid? The answer changes everything.
The shortcut that doesn't really work
A lot of business owners try to solve this by writing longer prompts. They paste in three paragraphs of background every time. That works the first few times, then it becomes a chore, and the AI tool falls into "useful for one-offs, not for the way I actually run my business."
The longer-term move is to give your AI a persistent place to read context from. A voice doc that lives in a shared drive. A customer record that gets pulled in automatically. A Foundation of how your company actually runs.
We wrote more about what that looks like in The Core, which is the version of this we're piloting with a small number of clients. For now, the main thing to know is this: when an AI tool feels disappointing, the answer is almost never "switch tools." It's "give it more context."
Why the model matters less than the context
A question we hear in every workshop is which AI to buy. Fair question, wrong emphasis. The big models leapfrog each other every few months, and whichever one you pick today gets outclassed by somebody's release next quarter. Your context outlives all of them.
Programmers have a useful way of seeing this. In software, the codebase is the asset. Programmers come and go, editors change, and the company survives because the code stays. An AI setup works the same way. The model is the programmer, hired labor you can swap. The written-down version of your business is the codebase, and it's the only part that compounds.
Big companies are starting to organize around the same idea. Block's "From Hierarchy to Intelligence" (Jack Dorsey and Roelof Botha, March 2026) calls it a company world model. One machine-readable record of how the business works, read by every person and every AI tool. Their version coordinates thousands of employees. Yours can be a folder of plain files, and the payoff shows up in weeks.
This is also why we build everything to be portable. Your context lives in plain files you own, and they read the same in Claude, ChatGPT, or whatever ships next. Switch tools, switch consultants, switch your mind. The asset stays yours, and nobody gets to hold it hostage. Not the AI companies, and not us.
What to try this week
Pick one thing you've already asked ChatGPT or Claude to do. A follow-up email, a job description, a customer FAQ.
Before you ask it again, write three sentences:
- What does your business do, and what's your tone?
- Who is this for?
- What's the actual goal of the message?
Paste those at the top of your next prompt. The difference is going to be obvious.