Igor Pogany tests ChatGPT Work, Claude Cowork and Gemini Spark on meeting action plans, business-document questions, customer-feedback analysis and inbox triage. The document and feedback examples use prepared data based on recurring business needs, not a standardized benchmark. He compares the outputs against the requested brief as well as his own preferences.
Igor Pogany highlights ChatGPT Work's distinction between confirmed and implied action ownership in the meeting test and its concise, cited practical answers to document questions. He often prefers Claude Cowork's readable writing and actionable synthesis, but notes that it sometimes assumes details or omits citations. In the inbox example, he personally favors Claude Cowork's filtering despite the model-based evaluation preferring ChatGPT Work.
Igor Pogany explains that his initial evaluator was an OpenAI model and checks its verdict with Fable, which also favors ChatGPT Work overall. Agreement between model judges does not establish an independent or exhaustive ranking. His practical conclusion is to combine grounded, instruction-following output with strong narrative synthesis according to the task rather than relying on one universal winner.
Watch on YouTube
