The human-AI pairs were strong on the writing and weaker on the images, though once the campaigns actually ran on social media the results evened out. The finding that stuck with me was about personality. The researchers matched people and AI on the Big Five traits, and the match changed the quality of the work. A conscientious person paired with an open AI produced better images. An extroverted person paired with a conscientious AI produced worse text and images. Even gender and cultural differences moved what the best pairing was.

The lesson for a business is simpler than the replace-or-augment debate everyone is having. Dropping an agent next to a team and hoping is the wrong way to do it. The result depends on how the work and the people around the agent are arranged, and that is the part most rollouts skip entirely.

It is why we spend the first week of any build on the people and the hand-offs, not the tool. An agent that fits how your team already works tends to stick around. The ones dropped in without that thought get quietly abandoned within a couple of months. Finding the place an agent actually fits is the whole job of an AI Audit, and it is how a single workflow grows into a Unified Agentic AI System that your team trusts.

Source: MIT Initiative on the Digital Economy, four studies on agentic AI, 2026.