What makes artificial agents intelligent? We perform group thought experiments where we assume that individual technical or resource obstacles can be removed. What if we could find (or build) vast amounts of marked conversational data and “remember” all individual user-agent interactions over all time? What if speech synthesis could be as subtle and nuanced as human speech? What if speech recognition worked at human levels of accuracy? What if we could add some new magical component? Where is the biggest gap in today’s intelligent, conversational, synthetic agent platform?