Test a browser agent on everyday sales research and scheduling

Grant Harvey and The Neuron team ran OpenAI's Operator through real sales tasks and reported what held up. Competitor pricing research worked best: it visited pricing pages, built a comparison table and asked for help at barriers before carrying on. Lead research was harder. After they took over to solve a CAPTCHA, it spent 15 minutes and came back with generic company profiles. The fix was to restate the source, the goal and the companies already found each time control returned. Case study summaries and scheduling across time zones also went well. Their verdict was that Operator is too slow for hands-off prospecting at scale and costs 200 dollars a month with ChatGPT Pro. The issue carries sponsor ads.