[DigitalToday reporter Chi-gyu Hwang] "Many companies buy AI licences and even run pilot projects, but their way of working stays the same. On the surface nothing looks different, but they may already be falling behind."

Mark Eisenstadt (마크 아이젠슈타트), founder of Limestone Digital, a company that places AI engineers at portfolio companies backed by private equity funds, recently stressed on social media platform X (Twitter) that companies delivering results with AI have changed their workflows.

Eisenstadt cited several survey findings and pointed to the lack of results that match the time and money companies have put into adopting AI.

A PwC report released in January said 56 percent of CEOs responded that they had not seen clear financial performance from AI. A McKinsey survey found that 8 in 10 respondents said productivity had improved, but only 37 percent of companies said it helped profits. That means even if employees finish work faster, it does not translate into company performance.

Eisenstadt explained the situation by citing the early days of electricity entering factories. At the start of electrification, factory owners swapped steam engines for electric motors but left line shafts and belts, which transmit power, in place. As a result, productivity hardly rose. Research by economist Paul David found that results came only after factories installed a motor for each machine and redesigned plant layouts.

The same applies if AI is introduced without changing anything else.

Eisenstadt said, "If processing a single request still goes through five departments and you are still waiting on an Excel file managed by one employee who is on leave, it is hard for an AI model to help no matter how fast it is. It is like attaching a new motor to an old line shaft and asking why the factory has not changed."

He said results differed for companies that changed their workflows.

In the McKinsey survey, nearly 3 out of 4 high-performing AI companies rebuilt workflows from scratch. Microsoft simplified its cloud supply chain work and then deployed about 100 dedicated agents, cutting average turnaround time to below 2.5 days from about 10 business days.

Eisenstadt also shared seven steps that lagging companies go through.

The first is a stage where the fact that the company bought AI is reported as if it were a result. The second is when employees use AI without approval. The third is when the finance team can explain spending but cannot explain revenue. The fourth is when it becomes harder to retain the talent needed. The fifth is when customers experience other companies' services and their expectations rise. The sixth is when the gap shows up in margins. The last is when the board demands an explanation.

Having technology does not necessarily lead to success.

Eisenstadt said, "Sears teamed up with IBM and CBS in 1984 to develop the online service Prodigy. Prodigy already supported shopping and flight reservations in 1990, 5 years before the Amazon website opened. Sears saw the future first but failed. Even if you recognise an opportunity and invest, it is useless unless you change the organisation."

Eisenstadt advised companies to start small if they want to change how they work with AI. Companies do not change in 3 months. If you change one task and prove results, the next task becomes easier.

He said, "If you cannot prove usefulness within 6 weeks, you should stop and examine why before putting in more budget," and added, "Do not believe the claim that you made that much money by multiplying time saved by hourly wages. You should be able to explain where the saved time went, whether costs fell, or whether output increased."