Paid 30-minute research interview: AI agent failures in finance and ERP workflows

Reasonwork AI is seeking experienced accounting, finance operations, ERP, or planning practitioners for a paid 30-minute research interview. We want to understand where frontier AI agents fail in real day-to-day work—not test you on a workflow or ask you to perform unpaid work. We are especially interested in firsthand use of ChatGPT or OpenAI agent workflows using GPT-6 Astra or GPT-5.6 Sol, and/or Claude Cowork with Fable 5.1 or Opus 5. Please tell us which products and model versions you have actually used; experience with other versions is useful context, but do not claim use of a named model if you have not used it. Relevant workflows include AP/AR, GL, reconciliations, month-end close, reporting, financial planning, and ERP work. Experience with one or more of NetSuite, SAP S/4HANA, Oracle Fusion, BlackLine, Workiva, Anaplan, Kyriba, Tipalti, Google Sheets, Outlook/Teams, SharePoint, or Tableau is useful; no one needs to know all of them. In the interview, we will ask for specific anonymized examples: what you asked the agent to do, what went wrong or was unreliable, how you noticed, what controls or workarounds you used, and whether the failure cost time or created risk. Please do not share confidential employer/client information, credentials, or live system access. Compensation: $50 fixed price per participant for one completed 30-minute interview. All pre-contract communication will stay on Upwork; the paid interview will occur after the contract is set up. When responding, please briefly state (1) your accounting/finance role, (2) the relevant systems you use, (3) the AI products and model versions you have personally used in that work, and (4) one example of a real failure or limitation you have observed.