You use AI. So do I.

The conversation has moved well past “should we use it?” and into something far more important: what kind of thinking are we outsourcing in the process?

I use Perplexity to pressure-test data and pull together research pathways fast. NotebookLM helps me work through dense reports and policy documents. Claude is useful because it pushes back on my thinking rather than simply agreeing with it. And ChatGPT or Gemini are often where I start when I need to get words moving on a page.

But I never ask AI to write a grant application for me.

Not because I think using AI is unethical or “cheating”. I don’t.

I think AI can be an incredibly valuable founder tool when it’s used properly. The issue is that too many people are asking it to replace strategic thinking instead of strengthen it. That’s where applications start to fall apart.

Most AI-written grant applications are not obviously terrible. In fact, many of them sound polished, articulate, and structurally fine at first glance.

What they often lack is judgement.

Nobody stopped to ask:
Is this actually true?
Is this the strongest argument available?
Would this survive scrutiny from an assessor?
What evidence is missing?
Does this genuinely align with the objectives of the program?

That’s the problem.

Across many applications, assessors are not simply evaluating whether something is well written. They are evaluating whether the founder understands the problem they are solving, whether the project is commercially and operationally credible, and whether the proposed outcomes actually make sense.

AI is very good at producing language that sounds convincing.

It is far less reliable at understanding whether your claims are realistic, strategically aligned, or supported by evidence. And honestly, assessors are becoming increasingly aware of the difference.

Not because they can always “detect AI”, but because shallow thinking leaves patterns. You start seeing vague impact claims, inflated projections, generic market positioning, and beautifully written responses sitting on top of weak commercial logic.

The strongest founders I see using AI are approaching it very differently.

They are not using it to replace expertise. They are using it to refine and pressure-test their own thinking. They use it to challenge assumptions, improve clarity, strengthen structure, and identify gaps before submission. The founder remains the strategist.

AI simply becomes part of the process.

Strong applications are not built by people who know how to write clever prompts. They are built by founders who deeply understand their business, the problem they solve, and why their work aligns with the objectives of the funder.

AI can absolutely accelerate the process.

But it cannot replace strategic judgement.