There is a concept gaining serious traction in AI development called context engineering. The Shopify CEO Tobias Lütke put it simply: can you state a problem so completely — with so much surrounding information — that a capable system can solve it without needing to gather more context?
I think this idea contains something profound. Not just for working with AI — but for understanding why so many neurodivergent people struggle in workplaces that were never designed for how their minds work.
The Narrow Context Window
In AI systems, a "context window" is the amount of information a model can actively hold and work with at any one time. Flood it with too much, and performance degrades. The system starts producing errors, missing nuance, losing track of priorities.
Neurodivergent people — those with ADHD, autism, dyslexia, and related profiles — often have a narrower working context window than neurotypical colleagues. This is not a character flaw. It is a cognitive reality. And it means that when the information environment is noisy, implicit, or hard to navigate, they hit a wall.
I experienced this directly. I used to work late evenings because the office was too stimulating in the afternoon — too many conversations, too much unpredictable input flooding my context window. The evenings were when I could actually think. But then I'd hit a wall: I needed to know how to use a university system, or check if I was allowed to proceed with something, and there were no guidelines written anywhere I could find.
What AI Is Teaching Us About Human Communication
The new science of prompting breaks down into four disciplines: prompt craft, context engineering, intent engineering, and specification engineering. Each one is essentially about eliminating ambiguity and making implicit information explicit.
The best managers have always done this intuitively. They give clear instructions. They define what "done" looks like. They explain the decision boundaries — what you can decide yourself, what you need to escalate. They make the rules findable.
Most organisations run on shared context that was never made explicit. Unwritten rules. Implicit hierarchies. Tribal knowledge held in the heads of long-serving staff. For neurotypical employees, this is merely suboptimal. For neurodivergent employees, it is often a barrier to functioning at all.
Context Engineering as Inclusion
Here is the reframe that AI is giving us: building a good environment for neurodivergent staff is the same problem as building a good context architecture for an intelligent system. You need:
- Explicit procedures that are genuinely easy to find
- Clear intent — what are we trying to achieve, and why does it matter
- Defined decision boundaries — what can be decided independently, what needs escalation
- Unambiguous instructions that do not rely on unstated shared assumptions
This is not a special accommodation. This is just good organisational design.