AI Unplugged: Is AI Coming for Your Job?

Like all disruptive technologies, #AI is about to blow open the job market as it affects white collar workers for the first time. So what can we do about it?

Better Than Nothing

A study of 774 job categories indicates that among the top 100, the most affected by language-based AI will include teachers (#2 through #9 and all throughout to #97), PR specialists (#34), and technical writers (#90). It's even worse if you're a teacher specializing in English or Communications, placing #2 and #13, respectively. In case you're wondering, writers and authors are #138 ... and the #1 spot goes to telemarketers.

In some ways, the possibilities of AI replacing existing jobs may not be a bad thing. Telemarketing jobs, which can often be awful for both the person working the job and the customer, could make the whole experience more pleasant while managing the workload better.

Klarna, for example, is replacing the equivalent of 700 full-time human agents with AI agents. But before we get too excited (Klarna certainly is!), it's worth taking a moment to remember that's 700 jobs that didn't enter the labor market. And if you need convincing on how much AI can replace telemarketers, just try the demo of Hume's Empathic Voice Interface (EVI). Talking to him for five minutes will make you a believer that any job relying on voice communication will be threatened by AI (and that's the model Hume's selling).

Nevertheless, there's an argument to be made that some industries perform poorly with humans in their roles, and efficiencies gained might be a net benefit if the whole industry performs better. This is the theory of Comparative Advantage:

The principle of comparative advantage says that whether the jobs of the future pay better or worse than the jobs of today depends to some degree on whether AI’s skill set is very similar to humans, or complementary and different. If AI simply does things differently than humans do, then the complementarity will make humans more valuable and will raise wages.?

Automation doesn't always benefit everyone equally, however. And for an illustrative parallel, let's look at another form of ubiquitous automation, the self-checkout in the United States.

Check Yourself Out

Self-checkout systems are just what they sound like: customers check themselves out, scanning their own items at a scanner and paying for the products, automating several processes that would normally be handled by a human cashier (checking to see if the number of products are accurate, tallying up coupons, running the credit card or cash transaction, etc.).

The wave of self-checkout systems resulted in cost savings to retailers. During the pandemic, self-checkout systems were particularly appealing because there was less human contact, with use increasing by up to 30% of transactions (almost double that from 2018). Despite this, the number of cashiers has remained largely unchanged for the last decade. At a glance, it would seem the theory of Comparative Advantage is in full force. Or is it?

The Bureau of Labor Statistics predicts a 10% decline in cashiers by 2032. Not surprisingly, the largest cashiers union, United Food and Commercial Workers (UFCW), which represents about 1.3 million workers in the U.S. and Canada, is against self-checkout machines.

But it's a little more complicated than "self-checkout machines destroy jobs." Alexandra Mateescu , an ethnographer and researcher at Data & Society, and a co-author, with Madeleine Clare Elish, PhD , of “AI in Context: The Labor of Integrating New Technologies,” explains why:

While self-checkout did not kill the cashier job, it did shift the roles, responsibilities, and perceptions of frontline retail workers’ jobs in often detrimental ways ... Self-checkout machines have played a role in the casualization of retail employment by making it easier for stores to reduce total worker hours and to more heavily rely on part-time staff, because the machines can be made active or inactive on an as-needed basis ... the impact of these retail technologies ... facilitate cost-cutting measures such as relying more heavily on part-time employees, understaffing, and intensifying work activities. In this context, employers can place greater pressures on frontline workers to absorb the consequences of these business decisions.

In short, self-service technologies like self-checkout machines haven't reduced cashier jobs (as indicated by the unchanged number of cashiers in the U.S.) but rather reconfigured the skills and responsibilities expected of them. The real danger isn't in removing jobs, but in making them much worse -- so much so that society begins to devalue the work itself. Astra Taylor calls this "fauxtomation," where automation's advantages are touted as a net benefit to companies by slashing jobs, even if the system isn't up to the task. And so the cashier's job shifts to "technology minder" where a human has to make up for the invested technology's shortfalls.

Sound familiar?

The Vulnerable Jobs

Cashier jobs are tough. Speaking from personal experience, I lasted one day at a customer-facing job as a teenager, and happily went back to my job working in a perfume factory (it was THAT BAD). I would later sell used computers at a resale shop and had long days where I often couldn't feel my toes from standing all day, berated by hostile customers who suspected that salespeople (that is, me) were going to metaphorically (if not literally) rob them.

Just as cashier jobs were radically transformed by self-checkout systems, white collar jobs will similarly be impacted by AI. The question is, will it be comparative advantage (everyone wins!) or fauxtomation (everyone secretly loses)?

Thanks to the Internet, writing reigns supreme. But instead of English majors ruling the Internet with their language mastery, online discourse gradually degraded to memes rife with misspellings. In other words, it doesn't matter if you're highly proficient in a language, because for most folks, they don't perceive the value of "loose" vs. "lose." And to be fair, there's a lot of necessary writing we create every day that is terrible, from form letters to surface-level bios to marketing brochures. This type of content is what my high school English teacher called "scarf-and-barf," the act of absorbing and regurgitating information. AI does scarf-and-barf exceptionally well.

Thing is, we need this kind of writing. It's not pretty, but transactional writing makes the world go round. The danger in replacing human-generated transactional writing with AI, like the automation of cashier jobs, is that with this automation comes the disdain for the job itself. I didn't like being a cashier or in sales in a retail store, but that doesn't diminish the value of the work. What's happening now with AI is that this same kind of transformation is coming for teachers and writers. And it doesn't feel good.

Will we all be out of a job? Not necessarily. Will the job, in whatever capacity that might be, become more about monitoring an (AI) machine that has been sold as fauxtomation (like cashiers monitoring self-checkout machines)? Quite possibly.

Pulling Back the Mask

AI is hot right now because it acts like a genie: ask it a question and it gives you an answer. That doesn't mean it's always right. And because these genies are trained on the bulk of humanity's public-facing content on the Internet, it can be misinformed and (sometimes hilariously) even making spelling mistakes. For some short-sighted companies, AI is absolutely going to cause job losses.

In the creative field, this is a real threat. The Writers Guild of America (WGA) and the Screen Actors Guild-American Federation of Television and Radio Artists (SAG-AFTRA) have started negotiating with future employers on the terms of engagement with AI and their business, and it begins by defining what is fully AI-automated, what is AI-assisted, and what replicates a creative human's work. It's a good first step, but the AI industry is moving so fast that it's not enough. We'll likely need government intervention at some point to help establish and define standards that protect workers.

AI has a cost too, even if it's invisible to users. AI takes up enormous electrical power and generates considerable heat, both precious quantities that have to be balanced against human needs. We can't have AI do everything for the simple reason that we can't afford it all.

Perhaps most important, AI doesn't do everything well. Sure, ChatGPT can write your essay for you -- but you'd better check that it's accurate. And soon, there will be counter-AI agents whose job is to detect AI in industries where precision matters. It's really important that jet engines work perfectly; instruction and technical manuals will need to be very precise, and it will be a long time before anyone blindly trusts AI to create a jet engine technical manual without checking its work for accuracy. When facts matter, like in legal briefs, AI is increasingly coming up short -- and arguing that "hallucinations" are feature, not a bug, of AI is not a defense according to several court cases. In short automation, including AI, isn't a one-and-done tool where we just turn it on and let it run. It needs human minders ... just like self-checkout systems.

This is why self-checkout is being reconsidered. Retailers are rethinking the system, at least in part because of merchandise loss, or "shrinkage." According to the National Retail Federation (NRF) shrinkage is on the rise: In 2021, retail to $94.5 billion, up only 4% from 2020 but a 53% jump from 2019. But that doesn't necessarily mean it's due to shoplifting or lack of humans monitoring self-checkout. Quite the opposite: what's really happening is that the systems themselves are responsible for shrinkage, thanks to scan errors, payment errors, and a host of other technical issues. Or to put it another way, while retailers may claim crime is the cause of their losses, they're really reversing course because the self-checkout systems they invested in are proving unreliable. Cashiers and customers know this because they've been picking up the slack for these systems for years.

Will AI take your job? The real question we should ask is: will AI change my job? It already has. All we can do is determine if the tradeoffs are worth it.

Please Note: The views and opinions expressed here are solely my own and do not necessarily represent those of my employer or any other organization.