The 9 Professions at Greatest Risk of Automation
The rapid improvements in artificial intelligence are leading to significant changes in the job market. AI has the potential to not only replace many professionals directly, but also indirectly by significantly improving productivity and reducing the number of workers required to complete a given task.
While few jobs are truly safe from automation, some professions face much greater risk than others, especially in the short term. I will briefly cover some of those most at risk professions in this guide.
Note: For a full list of jobs in each industry and their automation risk, you can refer my guide to the AI Revolution.
What Makes a Job Susceptible to Automation?
Several factors increase a job’s likelihood of being automated:
- The majority of the work can be done on a phone or computer
- The work is highly repetitive and follows simple rules
- There are no regulations in place preventing the use of automation
- The generated output can easily be understood and verified by humans
Professions at the Greatest Risk of Automation
1. Accounting
AI is currently capable of automating many basic accounting tasks, including using optical character recognition software to automate repetitive data entry, and large language models to generate reports and perform basic data analysis.
Some accounting firms have been slow to adopt even existing automation tools, meaning that the rate of automation in the industry has been surprisingly slow. However, this is likely to change quickly as accounting software providers like Intuit are starting to increase the implementation of AI into their products.
On the bright side, certain accounting professions such as auditors may still be required for the foreseeable future to manually verify records even when AI achieves perfect accuracy, mainly due to legal and regulatory requirements.
2. Digital Marketing
Large language models like ChatGPT can quickly and affordably produce vast amounts of writing in any style for purposes such as website copy, social media posts, and other marketing content.Â
As a real world example, many businesses ranking at the top of Google already use AI-generated articles to rank their websites, and media outlets are using AI to summarize news events in record time.
AI can also automate many SEO tasks, including keyword optimization, link building, and email outreach campaigns. This automation capability poses a significant threat to content writers and those working in other digital marketing professions like social media management.
3. Online Customer Service
Businesses are increasingly utilizing chatbots to manage customer interactions on their websites and social media platforms. These chatbots use natural language understanding to comprehend customers’ queries and provide suitable responses, similar to human customer service agents.
While chatbots can sometimes be frustrating for users or produce misleading information, this is largely due to poor implementation rather than inherent limitations. As businesses learn how to use them more effectively these problems are likely to be reduced.
As text-to-speech technology improves and becomes more natural and affordable, voice agents will also begin to replace customer service agents in handling phone calls as well.
4. Translation
The translation industry has undergone significant automation in recent years as a result of AI translation software like DeepL. AI can instantly translate text across multiple languages, though it still struggles with context, casual language, sarcasm, metaphors, and implied meanings.
Currently, the majority of work in the field involves editing machine-generated translations. As language models become more accurate, the need for human translators will continue to decrease. This automation trend will likely also impact interpreters as improvements in voice recognition technology enable real-time translation without human intervention.
5. Editing
Large language models can quickly and affordably edit substantial amounts of text, improving grammar, spelling, and writing style while adjusting content for specific audiences. They can also analyze text for logical inconsistencies and unnecessary repetition, and maintain consistent style through training on writing samples.
The main limitations slowing automation in this field are difficulties with fact-checking and performing accurate research, and maintaining a consistent story and characters when it comes to creative writing. However, most writers will be able to perform the majority of this work themselves using AI to assist them.
6. Tutoring
Large language models make capable tutors across various fields due to their extensive training data. They are affordable and can answer questions 24/7 whenever you need assistance. While they sometimes provide inaccurate information and may struggle with advanced academic problems, these limitations are being addressed rapidly.
There are also other types of AI models that can eventually be integrated with LLMs to create more intelligent tutors for specific subjects. For example, Google’s AlphaGeometry 2 has shown significant improvement in solving olympiad-level mathematics problems.
The majority of school teachers are likely safer from automation compared to tutors due to the need for in-person presence, maintaining discipline, and student motivation.
7. Law
Large language models are capable of analyzing legal documents, conducting research, drafting contracts, and providing basic legal advice to individuals and businesses. They can process vast amounts of case law and legal precedents much faster than human lawyers.
Legal software providers like Lexisnexis are beginning to roll out AI features into their products to answer legal questions and make it easier to find relevant cases, which will significantly increase the adoption of artificial intelligence into law firms.
The legal experts currently facing the greatest risk of automation are those who assist lawyers with tasks like research, administration and document analysis, such as paralegals and secretaries.
8. Graphic Design
Generative AI tools like Midjourney have improved significantly in recent years, enabling users to instantly generate realistic images from scratch and make quick edits with features like background removal, generative fill, and upscaling.
The main limitations of these tools include difficulty maintaining consistent brand imagery/style and achieving the exact desired results. AI-generated images also often contain noticeable mistakes and strange artifacts, especially in photorealistic styles, meaning that a human touch-up is currently still required for important work.
Freelance artists serving small businesses and selling their own designs on platforms like Etsy face the highest automation risk, as their clients generally have smaller budgets and can tolerate minor imperfections.
9. Software Engineering
One of the most impressive capabilities of large language models is their ability to quickly generate functioning code. In addition, they can also translate between programming languages, analyze code for flaws, and assist with debugging.
While they often make significant errors and require professional oversight, the programming capabilities of large language models are rapidly improving. They are also being increasingly integrated into tools like GitHub Copilot and Cursor, significantly enhancing the productivity of programmers.
Eventually, large language models will be able to create working software with minimal human intervention. While human software engineers will likely still be required in the short to medium term to manually understand and verify AI-generated code, smaller teams will be able to accomplish what previously required large development teams.
