Jobs to be automated 😱 😱

Six months ago I wrote a post called AI is coming for your services job which argued that many lower-skilled roles were at risk of automation. This is certainly proving to be true, but having met hundreds of companies since then, I believe it vastly undersells AI’s impact. As models have improved, AI has become highly proficient at reading, researching, summarizing, writing, extracting, speaking and listening. With recent advances in reasoning models like DeepSeek R1 and OpenAI’s o-1, models have also learnt how to use logic when solving problems. What’s clear is that jobs across the spectrum will change drastically.

As an investor this is extremely exciting stuff (but as a human, it’s scary) as I believe the next 5 years provides an unprecedented opportunity to transform virtually every industry. At 1984 Ventures, we have already made labor automation/augmentation investments in categories like software engineering (Cline), medicine (OpenClinic), medical scribing (Deepscribe), recruiting (Apriora), sales prospecting (Bluebirds), debt collection (Collectwise) and AR collection (Stuut) - and are on the hunt for more.

Below, I share a running list of jobs and tasks whose core functions are already capable of being done by AI today. Caveat: I don’t think these jobs will completely disappear. In many cases, the end state may be copilot tools that augment humans. In others, it will mean that humans predominantly review and curate the outputs generated by AI. Overall though, substantially fewer humans will be hired into the role going forward.

Jobs that will be automated

I’ll keep adding to this list over time (let me know if there are jobs that I’ve missed or opportunities that you feel are mischaracterized).

What’s interesting is the breadth of jobs and skillsets. There are monotonous and repetitive jobs, creative jobs and also high skilled jobs. Anthropic’s recent report highlighted that to-date, “AI’s use has been most prevalent for tasks associated with mid-to-high wage occupations like computer programmers and data scientists, but is lower for both the lowest- and highest-paid roles.”

AI for experts

What’s interesting is that many high skilled professions that involve the delivery of expertise (e.g. doctors) are actually at risk of automation (or at least substantial augmentation). This is because the corpuses of knowledge that underpin an expertise ca be fed into an LLM as part of pre-training. This has already happened with software engineering. It wouldn’t surprise me to see AI doctors, therapists, nutritionists and tutors in the coming five years - particularly as reasoning models continue to improve. This will have the interesting effect of democratizing access to experts at a fraction of the cost, opening up markets that may have historically been priced out of such services.

Anthropic’s chart above shows % of conversations on Claude pertaining to a particular job (in orange) and the % of jobs in the US that correspond to that profession (black).

Tasks that will be automated

Automation can also be thought of by task rather than job, with a task being something done across a number of jobs. My view is that tasks focused on research, content generation and code generation are already being made radically more efficient by LLMs. Some examples below.

A recent quote from Goldman Sachs CEO David Solomon highlighting AI’s content generation capabilities.

Advice for founders building labor automation startups

Just because a job can be automated doesn’t mean it will be an easy sell. Companies are often skeptical about whether a piece of software can fully mimic their human employees - and even if they can, they are typically reticent to fire large swathes of their workforce. Additionally, there is an inherent tension in asking employees to learn to use a new software tool that is being designed to eventually replace them. Some pieces of advice for founders below:

• Revenue generation value propositions resonate better than cost cutting or time savings
• It’s easier to position your product as augmenting humans rather than replacing them - turning them into 10x employees. (Anthropic’s report shows about 60% of queries involve augmentation)
• If automation still makes the most sense for your use case, it’s typically easier to sell a solution that automates a task that wasn’t previously being done. Our portfolio company Revv has found success in their vertical (auto) by automating a research task that workers were previously not doing because it was painful and time consuming

• Another alternative is to start by automating a task, rather than a job. This approach has the more subtle impact of reducing the need to hire for the role going forward (once the company realizes how much more efficient their employees are with the software), rather than asking the company to fire the staff upfront.

(Tagline from a website recently showing a good understanding of this nuance)