In a May 21, 2026 interview with Casey Newton (Platformer), Box CEO Aaron Levie argued that AI tends to automate

Box CEO Aaron Levie told Casey Newton in an interview that AI typically takes over the first 80% of work tasks — heavy, repeatable data processing and preparation — while the last 20% contains the expertise and judgment that produce most of a profession’s value. Published May 21, 2026, the interview highlights why this pattern matters: automation can change which parts of jobs are vulnerable and which remain the core of paid professional work.

Levie spelled out what falls into that automated 80%: reading and matching documents, initial text processing, compiling information and other steps that can be trained. As he put it, "The extra 20%, it turns out, is all the value creation of that profession. All the expertise and domain knowledge is in that last 20%, not the text that got generated." That distinction, he argued, separates routine task work from the judgment — driven contributions clients actually pay for.

The interview used a law firm example to illustrate the split. Junior lawyers and paralegals spend much of the week doing the document review and case summarization that AI can handle — the so-called 80%—whereas clients pay for the final 20%: crafting arguments, shaping strategy and persuading a judge or salvaging a complex deal. In other words, what remains valuable is the human capacity to interpret context and make high-stakes decisions.

Levie and the article place these dynamics in a broader labor — market context: people's anxieties are often about losing specific tasks — writing code, researching documents, doing initial analysis — rather than entire professions evaporating. The implication for teams and hiring is that AI may replace routine stages or roles, shifting competition toward those who can add unique, contextual value that machines cannot reproduce.

The piece also offers a concrete skills takeaway for practitioners and product builders. It points to roles such as cybersecurity engineers who must make rapid, consequential decisions during an attack as examples of where human skill remains decisive. The recommendation is to invest in decision — making under uncertainty, deep domain expertise, relationship management and responsibility in critical situations — capabilities that preserve and extend the "last 20%." The publication’s closing message cautions against dismissing accumulated human experience: automating routine processes can free time for higher — value work, but professionals must deliberately refocus from execution to protecting and producing the nonmechanical final steps of their craft where real value is created.