The most common question we hear from executives right now is some version of “what are other companies doing with AI?” The honest answer is: not as much as you think.

Most organizations have AI on the roadmap and some licenses floating around. But when you ask what has actually changed in how the business operates day to day, the examples are thin. Impact is still minimal — and few companies have publicly announced improvements in either revenue or profitability tied to AI.

The gap between expectations and reality is caused by a problem in the foundation.

A term worth borrowing

In software engineering, there is a concept called “technical debt.” Ward Cunningham, one of the creators of the Agile methodology, coined the term in 1992 to describe what happens when teams ship fast but skip the cleanup: code that works today but becomes harder and more expensive to build on tomorrow. Every shortcut compounds. Every workaround accrues interest.

Most organizations carry a version of this that rarely gets named. Call it “operational debt” — the accumulated weight of manual workarounds, disconnected systems, tribal knowledge, and duct-taped processes that have kept the business running but now make it nearly impossible to build anything new on top.

AI did not create operational debt, but it is exposing it. This is because AI requires exactly the things operational debt destroys: clean data, documented processes, and systems that talk to each other.

What operational debt looks like up close

A CEO we spoke with recently stepped into a company a few months after a major system change. The team had flipped the switch all at once — no pilot, no workflow mapped out — and assumed they would figure out the rest as they went.

Sixty days in, they realized there was a step earlier in the process that had to happen before the company could bill for work that had already been done. Nobody had mapped how work moved through the new system, so nobody caught it. Two months of cash generation were broken, not because the technology failed, but because the workflow had never actually been designed. It had just accumulated over the years in the old system, and everyone assumed it would carry over.

Now, the same CEO wants to launch a new customer-facing initiative on top of that same foundation. The conversation keeps stalling on the same thing: there is no documented workflow for how customers come in, how they get routed to the right person, how payment gets captured before service, or how the teams involved pass work to each other. Everyone owns a piece. Nobody owns the full process map.

This is something we see all the time, and it’s a great example of operational debt. It’s not a technology problem but a process and knowledge problem that makes technology adoption nearly impossible. Said another way: AI cannot optimize a process your organization cannot clearly describe.

Sequence matters more than spend

The companies that are publicly talking about their AI and automation wins are those that have their foundations locked down.

Walmart, for example, has invested heavily in automated fulfillment centers and supply chain technology — but the company made these investments on top of years of operational discipline. They standardized processes, unified roles across channels, and built strong data infrastructure before layering automation on top.

The lesson applies well beyond retail. The organizations getting value from AI are not the ones that spend the most. They are the ones that did the boring, foundational work first.

Start with the foundation

If your organization is evaluating AI investments — and most are — the highest-ROI move might not be a new tool. It might be an honest assessment of the operational foundation you plan to build on.

Take, for example, a Fortune 500 company that had accumulated roughly 15,000 reports in a legacy reporting system over the years. Only about 4,000 were used regularly. Different teams pulled different numbers from different places, and no one had a unified view of the metrics that mattered. Before the company could do anything meaningful with AI or advanced analytics, it had to go function by function — finance, store operations, supply chain, merchandising — to identify which metrics actually drove decisions, standardize definitions across teams, and migrate everything to a modern data platform. This foundational work took more than a year. It’s not exciting, but it gave the organization a clean surface to build on.

This is not glamorous work. It does not make for exciting board presentations. But it is the difference between an AI investment that delivers returns and one that joins the growing pile of pilots that never scaled.

So why does operational debt persist when everyone can feel it? Because there is no clean ROI story for paying it down. Standardizing metrics, documenting workflows, connecting systems — none of that comes with a vendor pitch deck and a projected payoff. It is preventative work with an invisible return, which makes it nearly impossible to prioritize against an AI tool that promises measurable impact in 90 days. 

Yet that AI tool will not have the promised impact without a clean foundation to build on. By the end of 2025, 42% of companies had abandoned the majority of their AI initiatives — up from just 17% the year before. This is the result of organizations buying tools to put on top of a foundation that cannot support them.

This debt does not get paid down in a quarter. But like any form of debt, the interest accrues whether you acknowledge the balance or not. The companies that start now will be the ones that lead when everyone else is still shopping for tools without a foundation to run them on.