An AI Answer Should Not Create Another Research Job

AI can produce a summary in seconds. Checking it may take rather longer. Professional research takes review. But if the summary sends you hunting for every source and rebuilding every comparison, the stopwatch has stopped at a suspiciously convenient moment.

The useful measure is the whole job: finding, checking and correcting, not just generating. A fluent summary can still take longer to check than it took to write. For employment-law professionals and HR consultants, a hard-to-check answer can become one more document asking for attention.

Right now, we're figuring out what makes AI-assisted research worth a professional's time: a summary with enough source detail to check without starting the research again.

In an illustrative summary, two tribunal decisions both involve a “delayed response”. But which response, delayed when, and according to whom? Without that context, the researcher reopens both decisions, rebuilds the chronology and works out whether the comparison was useful at all. That short sentence has left someone with a fair amount of work.

A better starting point names the events, points to the supporting text and makes uncertainty visible. The researcher can then spend their time assessing the comparison instead of reconstructing it.

Tribuno's direction includes checking consistency and making uncertainty visible. Those checks help us look for weak outputs; they are not independent proof of accuracy or time saved. The product still has to earn its place on an ordinary working day, when the actual work is waiting.

When you assess AI-assisted research, include source-finding, verification and correction in the task. Notice what you had to redo before the result became useful. Tribuno should help reduce the faff around professional judgement. The judgement itself stays with the professional.