Skip to content
Jami Ojala
Notes
2026-04Operations6 minPublished

AI-assisted grant drafting without losing the narrative

How to integrate LLM drafting into a multi-year grant pipeline so the administrative overhead drops but the strategic story that actually wins funding gets sharper, not blander.


title: "AI-assisted grant drafting without losing the narrative" date: "2026-04" kicker: "Operations" summary: "How to integrate LLM drafting into a multi-year grant pipeline so the administrative overhead drops but the strategic story that actually wins funding gets sharper, not blander." status: "Published" readTime: "6 min"

Most grant writers who start using AI make the same mistake: they ask the model to write the grant. The result is competent, structured, and completely forgettable — the kind of application that scores 65/100 and disappears into the rejection pile.

The problem is not the tool. The problem is the prompt architecture. When you feed a language model your project plan and ask for a full application, it regresses to the mean of every grant it has ever seen. Safe verbs. Generic outcomes. Zero risk. That is the opposite of what a strong application needs.

The split-pipeline approach

What works instead is a split-pipeline approach. Phase one is human-only: define the strategic narrative, the theory of change, the one sentence that makes a funder lean forward. Phase two is AI-accelerated: turn that narrative into the correct structural format, generate the budget narrative, draft the evaluation framework, and produce the compliance annexes.

At Nuoret Kotkat I built this pipeline for STEA and municipal applications. The LLM handled the repetitive scaffolding — formatting, cross-referencing objectives to budget lines, generating indicator tables — while I owned the narrative spine. The result was faster drafting and stronger applications, because the human energy went into the argument, not the word count.

The specific stack

  • Claude for long-context drafting
  • A custom prompt library with examples of funded vs. rejected applications
  • A strict review checkpoint where no paragraph ships without a human coherence check

Speed without shortcut.

Where this goes wrong

The most common failure mode is skipping phase one entirely. Teams that feed their project plan directly into a language model get applications that read like project plans — because that is exactly what the model sees. The strategic argument, the one that makes a funder choose you over forty other applicants, requires human judgment that no model can replicate.

The second failure mode is over-trusting the output. LLM-generated text is fluent and confident. It sounds right. But it will fabricate impact metrics, invent partnership histories, and smooth over real risks with optimistic language. Every claim needs verification against actual data.

The principle

AI is a multiplier, not a replacement. It multiplies whatever you give it. If you give it a strong narrative, you get a faster application. If you give it a weak narrative, you get a faster weak application. The investment is in the narrative — the AI just handles the scaffolding.

Share

Want to discuss this? Write directly.

jami@impactnode.fi