Where it can be useful

Candidate-facing language tasks are a bounded starting point: explaining a role description in plainer language, helping someone compare published career directions, directing a visitor to an official application route or answering questions from a reviewed knowledge source.

That boundary matters. Risk management starts by deciding what the assistant may say, where each answer comes from, when it should express uncertainty and when it must hand an official process back to the candidate.

What it should not do

A careers assistant should not silently rank people, infer sensitive characteristics, promise an outcome, invent vacancies or turn a conversational answer into an application decision. A good interface makes the limits visible and links to the authoritative next step.

Questions to ask before introducing one

  • Which sources are reviewed, who owns them and how are changes reflected?
  • Can a candidate reach the source or official route behind an answer?
  • What happens when the assistant lacks current information or receives a sensitive question?
  • Is there an accessible, non-chat route to the same essential information?
  • What evidence will show that answers remain supported and useful for the intended task?
A useful first scope is not "automate recruitment". It is "help a visitor understand published careers information, with an obvious route to verify and continue".

Sources and further reading