Top 10 Surprisingly Endangered Jobs from AI
This is a series of 4 articles I am sharing here, for people who are concerned and eager to understand more about job displacement impact potentially caused by artificial intelligence technology. You would read about “safe” versus “endangered” jobs in this series. The jobs listed in each article are demonstrative from my research research and technological knowledge, which may or may not fit into your personal scenario.
I highly encourage readers to take those as references and inspirations, and to start re-imagine and re-strategize your career today for our shared future — powered by AI.
How to determine what jobs are safe/unsafe?
White collar:
- Repetition vs. strategic:
- Does your job have minimal repetition of tasks?
- Do you regularly come up with insights that are important to your company?
- Do you make key decisions that cross functions for your company?
- Simplicity vs. complexity:
- Do most decisions in your job require complexity or deliberation?
- In your job, do you need to regularly learn and understand a lot of complex information?
Blue collar
- Dexterity vs. repetition:
- Does it require at least a year of training to be qualified for your job?
- Does your job involve very little repetition of the same task(s)?
- Fixed vs. unstructured environment
- Is your job usually performed in different environments each time? (e.g., a taxi driver would always work in the same taxi)
- Is your work environment unstructured?
For all jobs: human-contact / empathy / compassion
- Is communication and persuasion one of the most important parts of your job?
- Do you spend >30% of your work time with people who are not employed by your company (e.g., customers, potential customers, partners) ?
- Is a key part of your job performance measured by how well you interact with people?
- Does your job result in happiness, safety, or health of those your directly service?
- Do you lead or manage people in your job?
Top 10 Surprisingly Endangered Jobs from AI
Sales and Marketing Research
Doing research seems like a complex job that is hard to replace, but “research” is an overloaded word. When applied to sales and marketing research (or any data research in general), it is more about sifting through vast volumes of data and producing insight, which is increasingly more difficult for humans to manage so much data, dimensionality, and complexity. McKinsey classifies these professionals as “highly susceptible” to automation.
Insurance claim adjuster
Insurance claim agents are flooded with moderate monetary claims, but to properly handle each one, he or she needs to sift through a huge amount of data, and deal with much uncertainty. Insurance companies often only randomly checks claims, or automatically reimburse smaller claims. With AI, all claims can be checked against historical data, to reduce fraud significantly, and have much stronger numerical basis for adjustments. Also note that insurance as an industry is a large numbers game, with significant profit and overhead, and a lack of transparency and information-symmetry. It is a prime target for disruption, using AI and a brand new approach (such as Lemonade in U.S. or WaterDrop in China).
Security guards
There are already some “security robots” (such as Knightscope) being deployed in offices and fixed environments. But the much more cost-effective version will be many cameras connected to an AI system that performs monitoring. Both types can include not only cameras and microphones, but also depth sensors, odor detectors, and thermal imaging systems. These sensor output will go into AI to check for intrusion (even in pitch-dark, fire, or gas leak that are hazardous to human). Both types will be installed still with some human supervision (on premises for larger spaces, or in response centers for smaller spaces), with the bulk of the positions can be eliminated. There will always be at least one security guard on-site, to approach people, decide on tough situations, and manage the AI system as needed. This will be a well-paid job, so security guards can learn to become an expert user of the AI system to retain his or her job.
Truck drivers
There is a shortage of truck drivers, so for the short-term the employment will go up. However, because trucks are mostly driven on highways, and highways are the easiest scenario for autonomous vehicles, truck drivers will be among the first to be displaced once the technology is sufficient (around 3–5 years). Labor unions and lobbyists may protect the jobs a bit longer in some countries, but the technologies will launch in other countries. Countries that launch sooner may benefit from improving autonomous trucking technology as AI gets better, and would need to address the displacement issue sooner.
Consumer loan underwriter
AI works well when there is a large amount of data, a simple decision, and very accurate labeling. Loans are exactly in this category — banks have a lot of loan history, the decision is simply yes or no, and the labeling can be based whether each user in the loan history defaulted or not. An AI loan underwriter can consider much more information to decide about whether to approve a loan. We’ve already seen commercialized AI programs that have a lower default rate than human loan officers, while maintaining the same approval rate.
Financial and sports journalists
While AI cannot write a New York Times column or a New Yorker masterpiece, most of the news are repetitive description of very similar events (such as quarterly earnings report of a listed company, or result from a baseball game). This can be covered by AI, and even include multi-angle videos with voice over. Also, AI can come up with titles that are appealing based on historic click-through data (for the editor to approve and adapt). Journalists should soon be learning how to use AI for standardized reporting, and cultivate their other reporting skills that are not as easy to be replaced.
Bookkeepers & Financial Analysts
Automation will play an important role in the future of accounting, and bookkeeping will be one of the first roles impacted. Automatic data entry and reconciliation is already taking over more of the bookkeeping task, with AI spotting patterns to helping business owners gain insight. The future of accounting will be dependent on businesses’ need for a human interface for information. People employed in this area should move from transactional and repetitive analytical to insight-driven work. Building trust and having deep understanding of the business owners (not just the businesses) will be the least replaceable work in accounting.
Fruit pickers
Many jobs in agriculture are repetitive jobs. For example, tractor driving, seeding, weeding, monitoring, and fruit picking. Fruit picking is not a desirable job, due to the lower pay, and the poor ergonomics. There are several start-ups building fruit-picking robots with enhanced computer vision and robotics. Agriculture robotics can help take over jobs that humans no longer want in wealthier countries. Once adoption drives down the price, AI could automate food production, and help eradicate hunger from the world (along with that, farm jobs). Drone companies like XAG is rolling out agriculture focused hardware and software integrated solutions.
Investment professionals
In the financial industry, automation wave first hit the commodities trading pit. Many types of investments involve either digestion of massive amount of information, or require incredibly fast decision-making. Both types are suitable for AI. Quantitative trading, personalized robo-advisors, and buy-side equity with greater reliance on big data and AI for actively managed mutual funds are some examples. Clearly, there are still many top-level investment positions remaining, in M&A, angel investing, and bespoke types of credits. But the number of highly-paid people affected in the next 10 years will be substantial.
Radiologist
AI job displacement isn’t just limited to lower-paying jobs. Radiologists in New York make $470,000 a year. I’ve used radiologist as my example for several years. Several AI scientists already demonstrated human-level performance for recognition specific types of cancers (melanoma, lung cancer) in X-Ray, MRI, or CT. Another company has demonstrated an analysis of blood flow through the heart 180 times faster than human. During the 2020 coronavirus global pandemic, AI was trained on CT lung scans to help accelerate doctors’ diagnosis for COVID-19 infection among patients. While it will take a while for AI to take over most of a radiologist’s job, this is a definite job to avoid, if you’re thinking of medical school.
