Agent workflows · 2026-09-22
Using an AI API to write a ÜNİDES project report
After settling on a project idea, this covers which sections of a ÜNİDES application report an AI API is actually useful for, and which section should stay entirely your own.
From idea to report: a different stage, a different use
The ÜNİDES AI project ideas article focuses on generating ideas tied to real needs across 11 policy areas. Once an idea is settled, the next step is putting that idea into an application report reviewers will understand; that's a different writing task, and an AI API can be put to use differently for it.
In report writing, the model doesn't generate the idea for you; it helps you express the idea you've already clarified more clearly.
Which section benefits from support, which doesn't
Drafting the problem statement section, summarizing the same idea at different lengths (a short summary, a longer narrative), or reviewing the report's language and consistency are tasks a model handles well; in these sections, the model saves time by offering a first draft or an editing suggestion.
The sections covering the project's original contribution, personal motivation, or concrete implementation plan, on the other hand, should rest entirely on your own experience and idea; using a generic phrase the model produced as-is in these sections weakens the exact thing that sets your report apart from other applications.
- Drafting, summarizing, language editing: can benefit from model support.
- Original contribution, personal motivation, concrete implementation plan: should stay entirely your own.
- Rewrite a model-produced phrase in your own experience rather than using it as-is.
What the reviewer sees is the final text
A reviewer reading the report doesn't know which sentence came from the model and which from you; responsibility always stays with you. Sending a model-assisted draft without verifying and editing it against your own knowledge and experience carries the risk of the report containing an inconsistent statement, or one that doesn't reflect your actual knowledge.
Frequently asked questions
Is it acceptable to submit an AI-written report as-is?
Not recommended. The model can offer a draft or an editing suggestion, but the final text should accurately reflect your own knowledge and experience; a draft submitted as-is doesn't guarantee that consistency.
Which section saves the most time with model help?
Tasks like drafting the problem statement, summarizing, and language editing save the most time. Model support offers limited benefit in the original-contribution and personal-motivation sections because those should rest on your own experience.