Don't Surrender Standards: Who Decides Whether AI Work Is Any Good?
Marcus Torres is the Chief Product Officer at Quickbase.
gettyOne of the most overlooked aspects of AI’s impact on organizations is how it will affect standards of excellence. While there’s a lot of talk right now about how AI might replace skills or create efficiencies, the real danger for companies of all sizes is mistaking the approval of AI work for judgment of quality.
AI does more than just change who does what work. It changes who develops the judgment to know whether that work is good.
Here’s the pattern to look for: As AI takes on more work doing the building—whether that’s code, copy, analysis or the first draft of nearly everything—more people are quietly sliding from the role of creator to approver.
This can often feel like progress, freeing team members from menial tasks. But it’s a double-edged sword if you’re not careful—because those menial tasks mean something. They can create or deepen human judgment, which stops people from becoming irrelevant. Or, at the very least, those tasks are what allow people to build a sense of what’s good and what’s not.
In other words, many organizations are still asking, “Can AI do this work?” Yes, of course it can, and it can do that work increasingly well. But there’s a more important question that’s not getting asked as frequently: Who decides if the work is any good?
You might say that there are people who check the work that AI does, and they’re the ones who decide if it’s good. That’s true, but only to a point. Developer surveys are already showing just how far that goes. The reality is that trust in what AI has generated, like code accuracy, has fallen, even as AI usage has steadily climbed. People are approving AI, even as they either believe it less or, worse, question it less. That’s not oversight.
And the slippage matters everywhere, not just on the major items. It could be approving an AI-generated client note that misses the nuances of the client’s concerns, accepting an AI-generated piece of analysis that gets key underlying assumptions wrong or leveraging a bit of code that works but can’t properly scale.
It also makes people passive in an environment where the imperative should be for them to be active. Organizations need human ingenuity and creativity more than ever, even (or maybe especially) in roles where AI has taken over some of the work. That discernment is invaluable, but it’s become something organizations now have to deliberately look for—a reversal from when these capabilities were simply assumed.
Companies and organizations should rethink that now—and maybe even start assigning it directly as part of people’s jobs. Standards ownership could mean that every team is responsible not only for using AI efficiently but also for defining what “good” looks like in their own context, whether that’s accuracy, usefulness, scalability or customer impact.
IT has long been the default for this kind of assessment, asked to hold the line for standards in many companies. But it’s often beyond their scope to judge it. IT is built to judge safety, governance, reliability and risk. It is not always built to judge whether a message is persuasive, whether an analysis answers the right question or whether a product decision meets a customer need.
Instead of asking IT to do a job that was never theirs, companies should use governed environments that set clear guardrails from the start. That gives workers more room to create useful, specific solutions while keeping those quality standards closest to the people who understand the work and the problems that need solving.
Start by asking yourself a few questions: Does your AI wireframe or feature spec reflect your user research, competitive differentiation or tech constraints that your team understands? Does AI-generated content answer specific questions for your customers or audience, or does it just sound polished? Does an AI-assisted cost model account for real business constraints like supplier relationships or seasonal patterns?
We have reached yet another turning point with AI that goes beyond a single team or tool. Everyone has spent the last few years chasing the question of whether they could build. That’s now been answered: yes. The harder question now is whether what gets built actually changes how the business runs for the better and raises the standard—or if it simply becomes yet another thing to approve.
Every organization is at this juncture now. One path with AI—an endless stream of approvals until there’s nothing left to approve—is passive. The other empowers people at all levels to decide what good means and make the tools work for that standard rather than around it. That’s harder work. But it’s more meaningful, and it requires the kind of nuanced human understanding of quality that AI can’t yet replace.
AI can always build something. That’s not the hard part anymore. Now it’s on us to decide what’s worth building and what standards we refuse to surrender.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
