Is There Really a ROI When Replacing Humans With AI?

Artificial intelligence (AI)’s impact on human labor continues to be the most significant cause for concern throughout the global workforce, with white-collar workers facing an extreme version of automation that has plagued blue-collar laborers for centuries.

But recent research is starting to show that initial fears of wholesale layoffs and fully intelligent working environments may be overblown and that humans still have a few advantages over even the smartest technologies.

The issue is largely the cost of AI training and development. While crafting an AI model to take on key functions such as data analysis and predictive maintenance is very much achievable, pushing this to a level that matches human judgment and intuition for a wide range of circumstances is prohibitively expensive in terms of resource consumption and data access.

Table of Contents Table of Contents

1. Key Takeaways
2. The Bakery Hypothesis
3. Augment, not Replace
4. Changing Roles with AI
5. Creative Control
6. The Bottom Line
7. References

Show Full Guide

Table of Contents

1. Key Takeaways
2. The Bakery Hypothesis
3. Augment, not Replace
Show Full Guide
4. Changing Roles with AI
5. Creative Control
6. The Bottom Line
7. References

The Bakery Hypothesis

For instance, a recent study by MIT posited a hypothetical bakery that might consider replacing its bakers with AI (PDF). Intelligent processes and robotics could automate the portioning and mixing of ingredients, the baking process itself, and even packaging and distribution. But bakers are also responsible for visually examining their products to see if they are of suitable quality. Theoretically, this can be done with image recognition and smart analytics, but at what cost?

The team reasoned that visual inspection might take up about 6 percent of a typical baker’s time, meaning a bakery with five bakers making US$48,000 per year each would save about US$14,000 in labor costs.

However, this is only a fraction of the cost of implementing a computerized visioning system and the intelligent algorithms needed to maintain reasonable quality control. It will take many years to recoup the cost of this investment, while the need to employ humans to produce a suitable product remains.

Augment, not Replace

The healthcare industry is perhaps at the forefront of AI deployment, with the technology making its way into processes ranging from drug development and genetic research to cost control and customer service.

However, a recent analysis by app developer Appinventiv estimates the cost of implementing a working AI platform can range from US$20,000 to over $1 million, and even the high-end systems are unlikely to enable wholesale replacement of fully trained healthcare employees.

At best, these systems can do what many experts have been claiming for years: take on the rote, repetitive tasks in areas like records management, patient monitoring, and even some surgical procedures, but only as a means to increase the productivity of health workers and enhance their value to their organizations. This does help the healthcare system become more productive and less error-prone, leading to broader treatment and better patient outcomes.

Changing Roles with AI

Even lists of jobs most likely to be replaced by AI are full of caveats that only some of the tasks that humans perform will be automated. The jobs themselves will continue, albeit in different forms.

In Business Insider’s more recent update, for instance, computer programming is one of the top jobs most likely to give ground to AI. But while large language models (LLMs) like ChatGPT can certainly churn out code faster than humans, they still lack the intuition to create truly innovative applications.

Again, this means it can take on much of the coding work, but the final product must be polished by an intelligence that knows exactly how it should function.

Creative Control

Likewise, human judgment is crucial for jobs like writing, legal services, marketing, and teaching, all of which stand to benefit from healthy doses of AI but are not full replacements. When it comes to the complexities of human interaction, as well as creativity and the ability to quickly respond to unforeseen events, the human brain is still without a peer.

The myth continues, however, that these qualities are only required in the upper echelons of the employment strata, when many “normal” jobs also rely on uniquely human traits.

Any building trade, for example, will remain in human hands for a long while, according to PwC’s Global Artificial Intelligence Study, which includes carpentry, electrical work, plumbing, and masonry. Hairdressers and stylists also have little to worry about, as do clergy, emergency responders, and musicians.

The Bottom Line

All of these roles will certainly benefit from AI around the edges. Still, their core responsibilities are simply too complex for algorithmic processing – and probably will be for a long time to come.

Technology, after all, is only useful when its benefits outweigh its costs, and at the moment, the human mind is still the most cost-effective option for the things that matter most.