What sectors are least vulnerable to AI or has the automation already happened?

We thought it would be interesting to ask ChatGPT the question, “which sectors are LEAST vulnerable to incursion by disruptive AI technologies?”

ChatGPT picked out some of the sectors you would expect, such as those requiring high human empathy, physical dexterity, regulatory complexity, or where there is low economic incentive for automation. Examples given include:

• In the manufacturing sector, high-precision, low-volume manufacturing of products such as aerospace components, custom medical devices, and luxury timepieces. Such products require craftsmanship, manual dexterity or costly tooling changes not easily automated or worth the investment;
• In the construction and skilled trades sector, specialist services such as plumbing, electrical work or masonry typically occur in complex, dynamic environments that require adaptive physical labour, on-site problem-solving and human judgment;
• Artisanal and craft-based manufacturing examples include handcrafted furniture, leather goods and gourmet foods. In these categories, buyers seek authenticity, imperfection or “human touch”, which AI or machines can’t replicate convincingly;
• Service areas that involve therapeutic and mental health services (such as psychologists, counsellors, social workers) are suggested as requiring deep emotional intelligence, trust and nuanced interpersonal engagement, all areas which AI still struggles to navigate effectively. Similarly, early childhood education and care services such as preschool teachers and nannies require physical presence, real-time emotional response and trust-based caregiving;
• In the legal services sector, many of the more “routine” areas such as conveyancing, wills or claims management are becoming increasingly automated, but high-stakes legal services (such as criminal defence or complex litigation) are suggested as requiring ethical nuance and a need for human judgment in adversarial and interpretive environments which might make full AI replacement unlikely; and
• Diplomacy and high-level negotiation; these activities require inputs such as awareness of historical background, labour mediation skills, tactical game-playing and often the exchange of “baskets” of differing value depending on the viewer’s background or point of view. Factors such as cultural context, strategic ambiguity, nonverbal cues and ethics may be relevant - all areas where AI has reduced capabilities

Common factors

ChatGPT suggests that AI’s current limitations are shown in areas that involve factors that are hard, if not impossible, to automate. Physical dexterity, emotional intelligence, contextual understanding, or common sense are typically viewed as particularly human. Others may require a human trust factor, where clients demand human accountability, for example where risk, care, or money is involved. Even where some element of automation is possible, it may not be economically practical, where the cost of replacing labour with AI is higher than the value AI can generate.

AI is not the same as automation

However, as we read through these lists, what struck us is that many of the sectors described have already been transformed, to a greater or lesser extent, by increased adoption of software or IP that automates a significant part of the process. Even where the service is required to be delivered by a skilled human, there has been a massive incursion of automation in the areas of learning the skills required, using rostering software to optimise the delivery by skilled people, using ERP or other accounting software to count the “score”, or at an earlier stage, using CRM software to win the business. In the manufacturing sector, more or less automated technologies such as robotics, CNC machining and 3D printing are now well established.

AI will eventually play a bigger role when it improves

Even in some of the areas highlighted above as being AI-resistant , AI is beginning to play an increasing role. For example, within the manufacturing sector, AI technology is playing an increasingly important part in generative design, motion control, process optimisation and quality control.

Even in the area of precision manufacturing of low-volume parts (the first example lighted on by ChatGPT as being resistant to AI), there is an interesting counter-example in Xometry, an online platform enabling you to “source all your CNC machining jobs, from prototyping to high-volume production, reliably in one place”. Xometry claims to offer massive capabilities with 10,000 manufacturers across Europe and globally and says that their “advanced AI technology, combined with attentive customer care, ensures a seamless experience from instant quotes to order tracking.”

To be clear, Xometry doesn’t actually own any CNC machines or carry out machining itself. However, its well-designed online platform enables customers to upload their product requirements efficiently and receive quotes from the bestsuited suppliers at any point in time. Those suppliers are probably not using anything close to their capacity, and can therefore quote at lower prices than normal, even for low-volume parts. In addition, the business-winning, and some of the design, elements of the process (which normally comprise a substantial part of the costs involved in winning a new customer) have been automated and informed by AI technology.

To plagiarise Marc Andreessen, perhaps software has already “eaten the world”. AI may be having another bite, but it is doing so by stealth, and sometimes in ways that are not immediately obvious.