October 6, 2026, (Inside AI) — As companies rush to integrate artificial intelligence into their workflows, a critical bottleneck is emerging: the challenge of verifying the quality of AI-generated outputs. According to MIT Sloan research scientist Christian Catalini, "When execution is cheap, verification becomes more valuable," a shift that should force companies to rethink where their value and risk truly lie.
Catalini argues that successful companies will become "verification factories: institutions capable of properly steering AI-generated output and standing behind the results." This transformation requires a nuanced understanding of where execution and verification costs are mismatched. In fields like coding, AI can make verification cheap and easy. However, in areas such as novel research and development, verification demands expert judgment, and the risk lies in either overwhelming experts or letting outputs go unchecked.
The stakes are high. Companies that fail to adapt may inadvertently erode their competitive edge by giving away two key assets: talent and unique ground truth. Catalini warns that many firms risk losing both in their haste to embrace AI and automation.
To avoid this, Catalini offers several pieces of advice. First, don't automate away your expertise. "Good managers are verifiers," he says. "Yes, they route information, but in doing so they also determine what it means, what deserves others' attention, and what meets their quality bar. In making these calls, they often retain a significant advantage even over frontier AI." If companies eliminate the friction that drives learning, he cautions, "they may not recognize the damage until the next big discontinuity, when the models are confidently wrong and no one inside the firm is still used to overruling them."
Second, be wary of unknowingly training the competition. While companies often focus on data privacy, how they verify quality can be just as crucial to their competitive advantage. Even with agreements that prompts and outputs won't be used for training, Catalini writes, AI systems might still capture "the distinctive ways employees handle exceptions, make decisions, and prioritize information. Taken together, those traces can provide a blueprint for replicating parts of a firm's verification engine."
These insights come amid a broader conversation about AI adoption. A study by BCG's Gabriella Rosen Kellerman, David Martin, and Julia Dhar of over 1,000 U.S. employees reveals that leadership narratives significantly impact AI usage. Framing AI as an expected part of the job predicted 40 more minutes of use per day, while emphasizing high-quality work predicted 18 more minutes. Fear-based messaging, such as saying AI is necessary to stay competitive, predicted 3 minutes less use. Ambiguity was worse than saying nothing at all.
The authors write: "AI transformations begin with our leaders' stories. Our data reveal that an important difference between companies with higher and lower levels of employee AI adoption and financial goal achievement is not budget, not tooling, not technical infrastructure. It is which story leadership chooses to tell, and how clearly they do so." They advise leaders to be crystal clear on expectations and treat ambiguity as a business risk.
Read: Only 25% of Leaders Confident in AI-Driven World, Survey Finds
As AI continues to reshape industries, the focus on verification and clear communication will separate winners from losers. Companies that invest in verification capabilities and craft compelling narratives will be better positioned to harness AI's potential while mitigating its risks.
