AI agents and technology
What is a context window? Definition and examples
Also: Context length
In short
The context window is the amount of text an AI model can read at once when it answers. Anything that does not fit, or is not provided, the model does not know at that moment. It is measured in tokens, roughly parts of words.
Why it matters: What sits in the context window decides whether an AI answer fits your situation.
Why does Context window matter for your company?
A model knows about your company only what is in the context window: the task, the conversation so far and the documents provided. Windows have grown a lot in recent years. But large does not mean good, because lots of unsorted text often makes answers less precise and costs time and money.
The real question is therefore not how much fits, but what belongs there. The valid decision, the relevant rules and the current state belong in it; old versions and side issues do not. This selection is the core of context engineering.
Example: The whole drive in the prompt
For a market analysis a team loads every strategy document from the last three years into an AI assistant. The context window is big enough. Yet the answer mixes old expansion plans with the current decision “Strategy 2027: focus on the DACH mid-market, no US expansion”.
On the second try the team provides only the valid decision, current market data and a clear question. The answer is shorter and fits the strategy. Less but correct context beats more context.
3 common mistakes to avoid
Thinking bigger is better
A large window invites you to pour everything in. The model then also weighs what is outdated and beside the point.
Letting the history grow forever
In long chats early instructions lose weight or drop out. Start new tasks fresh, with a short summary.
Mistaking the window for memory
What is in the window is gone after the conversation. What should last belongs in memory or a maintained source.
Step by step
- 1
Clarify the question first
A clear assignment determines which context is needed.
- 2
Provide only the valid state
Current decisions and rules instead of whole folders.
- 3
Name the sources
Anyone can then check what the answer is based on.
- 4
Restart long conversations
Summarise the result and start again from that summary.
Bring one important decision.
In 30 minutes with Marc or Simon we show where your work drifts from the decision today.
Meet the Company BrainRead on in the PM Lab
Articles that go deeper.
Related terms
- Context Engineering →Context engineering is the work of giving an AI model exactly the right context for a task: the goal, the rules that apply and the knowledge it needs.
- AI memory →AI memory is the part of an AI system that keeps knowledge beyond a single conversation.
- RAG →RAG (retrieval-augmented generation) is a method where an AI model looks up relevant passages in your documents before it answers.
- Hallucination →A hallucination is a statement by an AI that sounds convincing but is false or unsupported.