For the last few years, much of the conversation around generative AI has focused on prompts.
Prompt engineering. Prompt libraries. Prompt templates. Secret prompts.
Better prompts absolutely produce better results.
But we think the importance of the prompt is often overstated.
The system around the model is.
A prompt is a moment
Imagine asking an AI:
To do that well, it might need to understand:
You can put all of that into a giant prompt.
But should you have to?
And should you have to reconstruct it every time?
Companies accumulate knowledge
A great internal team doesn't start every project with amnesia.
Neither does a great agency.
Over time, people learn:
That knowledge improves future work.
AI should work the same way.
Context is infrastructure
Instead of thinking about AI as:
The more useful architecture is closer to:
Now the model isn't being asked to reconstruct the company from whatever happens to fit into today's conversation.
The system supplies the right context for the task.
And what happens during the task can improve the system.
Better models don't eliminate this need
Models will continue getting better. They'll reason better. They'll generate better design. They'll understand images better. They'll write better code.
That's great.
But a smarter model still doesn't automatically know your company.
It doesn't inherently know which direction your team approved last Tuesday.
It doesn't know why a particular positioning statement was rejected.
It doesn't know which previous creative work should become precedent.
Those aren't model-intelligence problems.
They're organizational-context problems.
The valuable layer is what surrounds the model
As foundation models become increasingly capable, differentiation moves toward the systems built around them.
Those questions matter enormously for creative work.
From prompting to creative intelligence
At Morphic, this is increasingly how we think about the category.
The goal isn't to create a magical prompt that produces the perfect website.
It's to build a system that understands enough about the company and the work to make every generation better informed.
The prompt still matters.
It's how a person expresses intent in the moment.
But the prompt shouldn't have to contain everything the system needs to know.
They were good because they understood you.
AI should aspire to the same standard.
