AI can help produce a policy brief, but producing fluent text is different from knowing whether its claims are supported. Employers introducing AI into junior analytical work should teach verification as part of the assignment. A final approval by a senior colleague is not enough if the writer never learns why a convincing sentence is wrong.
The learning problem begins after the draft
A junior analyst needs to understand which sources deserve weight, what a decision actually authorises and how uncertainty should be explained. A tool that drafts quickly can leave more time for those tasks, or make it easier to skip them.
The examples below are hypothetical. A fluent rewrite can change a request into completed funding, discretion into a promise, or an approval into money received. Verification means preserving what the source actually establishes, even when stronger wording sounds clearer.
| Status | Source says | Incorrect rewrite |
|---|---|---|
| Request | A proposal requests funding | The project has been funded |
| Discretion | The board may approve | The board will approve |
| Approval | Funding was approved | Funding was received |
AI research shows benefits and limits
A CHI 2025 study surveyed 319 knowledge workers about generative AI. Greater confidence in the tool was associated with less self-reported critical-thinking effort, while respondents also described work shifting towards checking and integrating output. The study reports associations and experience, not proof of permanent skill loss.1
Other workplace evidence shows benefits for learning. A 2025 study of an AI assistant used by 5,172 customer-support agents found larger gains among less experienced and lower-skilled workers, with evidence that some improvements persisted when assistance was unavailable.2
Neither study directly establishes the outcome for junior UN policy analysts. Together, they challenge a simple story in which AI either teaches everyone or prevents learning. The design of the work, the tool and the feedback deserves examination.
Make verification something the supervisor can review
Employers could test this approach through a supervised exercise. Give a junior colleague public documents covering a proposal, a decision and an implementation update. Ask for a short brief using an approved tool, then review the following steps. The exercise is a suggested teaching method, not a finding from the studies above.
- Question: Before drafting, identify what the brief must establish and which documents might answer it.
- Evidence: Trace each material claim to the original passage, checking dates, amounts and institutional terms.
- Correction: Explain any wording changed because it overstated or misread the source. A relevant-looking link is not enough.
- Understanding: Discuss what remains unknown and try a short unfamiliar example without the tool.
Keep responsibility visible
In the funding example, the learning is in explaining why a request is not an approval and why approval is not proof of receipt. If a supervisor silently repairs the sentence, the immediate document improves while the misunderstanding can remain.
A short record of claims, sources and corrections gives the supervisor and junior colleague something specific to discuss. The supervisor can ask why a sentence changed and check the colleague’s explanation, rather than judging only the polish of the draft.
These checks should be proportionate to the stakes. A routine internal summary and a public statement do not need identical review, but someone should know which claims matter and who is responsible for approving them.
Permission to use a tool is a separate question
The ILO Library's research guidance warns about unreliable AI information and the need to evaluate sources. It also cautions against entering confidential material or intellectual property without appropriate permission.3 Research guidance is not authorisation to use a particular service with workplace documents.
Staff should follow their own institution's current rules about tools, accounts and information. Training with public documents can teach source checking without assuming permission to upload internal material. Supervisors need to make that distinction clear before assigning the work.
Treat analytical judgment as part of the job
For a student entering international work, being able to explain a brief is a stronger aim than simply generating it. Which source supports the main claim? What has changed since it was published? Which sentence could lead a reader to infer more than the evidence shows?
The workforce questions raised by UN80 include how institutions will develop new professionals. AI makes that responsibility more pressing. Employers can use faster drafting while still teaching judgment, provided they assign the checking, review the reasoning and allow time to learn.
