Adding whiz-bang tech, the theory goes, can save businesses millions of dollars by making everything from manufacturing to operating retail stores more efficient. It also can help customers shop more easily, particularly online, thereby increasing sales.

But there’s a more complicated reality: Digital transformations are excruciatingly difficult to pull off. They’re expensive. And they don’t always achieve the ambitious business goals originally envisioned.

The trouble often boils down to one thing: human nature. Disagreements, poor planning, and delays are often to blame. “Everyone focuses on the digital and tech and not the human side of it,” Alex Holt, global head of technology, media, and telecommunications for business consulting firm KPMG, said on Tuesday at Fortune’s Brainstorm Tech conference in Half Moon Bay, Calif.

The trick is to create a company culture that makes digital transformations easier. As Atif Rafiq, CEO of Ritual, a newly announced startup that provides a system for managing agile work, described it, that effort is “a fight for the soul of the company.” That’s easier said than done.

Executives, for one, must be aligned in their thinking. They have to be willing to spend the money, provide enough employees to meet a reasonably fast deadline, and have a consistent goal in mind for what the project should achieve.

One significant danger is that chief financial officers looking to save money could derail a digital transformation midstream. So too could corporate lawyers, who can slow progress while they endlessly review its legal risks.  

Digital transformations also depend on technology, of course. And in many cases, they involve adding artificial intelligence to a company’s operations to analyze data so that leaders can make better decisions.

That depends on getting access to good data. And these days, many companies have more data than ever at their disposal as they increasingly digitize their operations. “Companies are sitting on this gold mine of data,” said Navrina Singh, CEO of Credo AI, a maker of software for managing the various risks involved in data.

But businesses often fail to make their information easy to tap. Data can also be obsolete or a duplicate of what’s held somewhere else by the company. And even if the data is ready, Singh said, companies sometimes neglect to vet it to determine whether it was collected ethically—i.e., after people gave their consent—or whether it could cause a product that depends on the data to be racially or otherwise biased, as some tech products have been in recent years.

Another reality check about digital transformations counters the widespread thinking—especially among companies without a long tech pedigree—that the process is finite, and that after finishing one, whether after six or 12 or 18 months, a company can simply relax and reap the rewards.

In fact, said Marianna Tessel, executive vice president and chief technology officer at tax software maker Intuit, a real digital transformation is never ending. There’s always more and better tech that a company can add, like a permanent home remodeling. “It’s continuous; you finish one thing and then there’s another thing and another thing,” Tessel said. “It’s not just starting from one point and getting to a destination.”

Correction (12/1/21): An earlier version of this article incorrectly described the focus of Ritual. It provides a system for managing agile work.