So what would happen if human CEOs were replaced with automated versions? The Hustle looked into that, noting that CEOs often do work that AI is strong at–budgeting, tracking a company’s performance, and making executive decisions based on data, that eliminating their very high pay could be a substantial corporate cost savings, and that even 49% of CEOs think most or all of their duties could be replaced by AI.

So they tested that proposition by producing an automated CEO for three companies aiming for revival: Nike, Southwest Airlines, and Starbucks.

This is what they found. While each company’s AI CEO showed individual strengths and weaknesses given each company’s circumstances, “Overall, the AI CEOs’ ideas weren’t much different than those of their human counterparts…” However, “the AI CEOs didn’t mirror their human counterparts on everything. When it came to topics like layoffs and customer relationships, the AI CEOs seemed to have a greater desire to do well by their employees and customers… but they lack the personal skills needed to inspire employees and get along with investors… {Also}, the tendency of AI to be too wishy-washy and easily influenced by whomever delivers a prompt is a common criticism. It’s also certainly not a good characteristic for CEOs, who must be decisive.”

So what’s the takeaway from this experiment? It’s interesting that the comparisons between human and AI CEOs focus on what are attributes of emotional intelligence–empathy for employees and customers, inspiring employees and getting along with investors, abilities that demonstrate emotional intelligence. Emotional intelligence has also been shown to elevate decisive decision making.

For lawyers and law firm managers, making some aspects of practice more efficient through the use of AI can be useful as long as the core advantages to a legal practice of emotional intelligence, like expressing empathy, inspiring the troops, successfully interfacing with stakeholders of all types, and making sound decisions, is kept top of mind–advantages that so far AI cannot reliably deliver.