Marriott operates more than 30 brands across 9,700 properties worldwide, the vast majority of them franchised. Each brand has extensive standards for guest experience and facility design, defining what the brand should look and function like in every location.
Now multiply that across industries. A national homebuilder delivering consistent quality across dozens of markets. A retailer maintaining brand expression in 500 storefronts. A university system trying to hold design standards steady across a campus that gets built out over decades by different architects, administrations, and budgets. These organizations have spent years codifying what makes them great.
And almost universally, that knowledge falls apart at the point of execution.
In 2008, Starbucks closed 600 U.S. stores and temporarily shut every remaining location to retrain baristas. In my opinion, it was because the brand expanded so aggressively that it drifted from its own operating standards.
I also think Target had one of America’s most refined retail playbooks, which fell apart when the company expanded into Canada due to higher prices and poor stock levels. In 2015, Target closed all 133 Canadian locations within two years of opening in 2013.
PwC’s consumer research shows 32% of customers will walk away from a brand they love after just one bad experience. In none of these cases was the problem a lack of knowing what great looked like. It was the gap between defining it and delivering it consistently at scale.
The instinct is to blame people. But the real issue is structural. Standards are written by experts in one context and executed by different people in another, under time pressure and with incomplete information. Every hand-off means something gets lost, simplified, or reinterpreted.
Across dozens of handoffs and thousands of decisions over years of execution, the gap between what was intended and what is actually built or delivered widens to an enormous extent.
AI as institutional memory
Most of the conversation about AI right now is about generation and making new things faster. But for organizations that have already done the hard work of defining what “great” looks like, the far more valuable application is fidelity. Making sure their definition of “great” survives execution.
Instead of asking AI to imagine something new, these organizations ask it to remember what was already decided and hold the line. AI trained on an organization’s accumulated knowledge becomes a kind of persistent institutional memory that travels with the work wherever it goes.
In brand management, tools like Adobe GenStudio now assign automated compliance scores to content in real time, checking voice, tone, and regulatory standards before anything goes out the door.
In architecture and design, we see it on nearly every project. A brand’s design standards can be embedded directly into the production workflow, so that every drawing is continuously checked by AI as it’s produced. A specified finish is swapped for a close-enough substitute, and AI flags it before it enters a drawing set. A detail that passes code in one jurisdiction but fails in another is caught by AI before it ships to the wrong market.
The pattern is the same across industries: AI absorbs accumulated knowledge that would normally reside in the heads of a few experts or be scattered across a few hundred pages of documentation, and ensures it travels with the work.
The paradox nobody expects
When teams spend most of their energy policing baseline standards, verifying, rechecking and defending what should already be settled, creative capacity gets eaten alive. When AI handles that fidelity layer, the people in the system get to operate at a higher level. They stop protecting the floor and start pushing the ceiling.
This pattern is not new. Toyota’s production system made it explicit decades ago. Taiichi Ohno put it simply: “Without standards, there can be no improvement.” He viewed standards as a floor, not a ceiling: a stable platform that let people see problems clearly and solve them creatively instead of spending their time figuring out what the baseline should be.
Toyota’s Creative Idea Suggestion System, running since 1951, has generated over 50 million improvement suggestions. In 2023 alone, employees submitted 810,000 ideas, about 14 per person per year.
In my opinion, the numbers suggest that rigid operational discipline enabled innovation.
That’s the opportunity AI creates for any organization with complex standards: Ensuring the institutional knowledge those humans spent years building actually survives every handoff between the people who set the standard and the people who carry it out.
A new divide forms
BCG’s September 2025 report found that only 5% of companies qualify as “future-built” for AI, while 60% report minimal gains. The difference isn’t the technology. It’s whether they’ve redesigned how work gets done or simply bolted AI onto old processes.
The competitive divide forming now is between those using AI to produce more and those using it to make sure what they’ve already defined actually reaches the world intact. Most companies are pouring AI into the front end, generating more content, more options, more output.
The ones that will be hardest to compete with are the ones that point it at a problem nobody talks about: Making sure the organization’s own hard-won knowledge doesn’t evaporate the moment it leaves the building.
Sandeep Ahuja is the CEO and cofounder of cove.