AI-Assisted Business Analysis: Usability, Trust Cues, and Human Oversight in Responses to Under-Specified Requests

Analysts increasingly turn to generative AI assistants for business-analysis work, yet they often start with short prompts that carry no organizational context. We examined how three general-purpose assistants respond to such under-specified requests and what their responses imply for usability, trust and human oversight. ChatGPT, Claude and Gemini each received five prompts that named only a common business-analysis task: requirements analysis, requirements prioritization, user-story generation, stakeholder-needs analysis and business-process improvement. Each prompt was run three times per system in fresh sessions on 24 and 25 September 2026, giving 45 outputs, which we coded for accuracy, completeness, internal consistency, usability, transparency and support for human oversight. ChatGPT and Gemini explained the named concept in all 30 of their outputs and never asked about the user's situation. Claude asked a clarifying question in 13 of its 15 outputs, and five of its outputs contained little or no task content. Gemini covered 59 of 60 checklist elements but accounted for five of the six accuracy issues. No output expressed uncertainty or suggested verification, and every substantive output was generic. The results point to a trade-off between immediate completeness and handing control back to the analyst. They support a workflow in which analysts supply context, check outputs and validate them with stakeholders before acting on them.