The quick answer to the question is $304,472.68 per year. That's how much it would cost to run a GenAI algorithm 8 hours a day, 5 days a week for 48 weeks out of the year. For most of people under that compensation threshold, you may be breathing a sigh of relief. GenAI is expensive to run! But it's also not quite that simple.
Make no mistake, we are heading into a period of significant job disruption. Much like the Industrial Revolution, machines are now capable of doing things humans once did. However, unlike the Industrial Revolution, we are increasing the complexity of work, not simplifying it. This means there will be fewer jobs overall, and those jobs will pay more to individuals with the skills to meet their demands.
The Good News
If that sounds bleak, don’t worry just yet. It’s worth doing a reality check on the current state of artificial intelligence and why it still faces challenges competing with (most) human jobs for now
- AI is task—not role—oriented
- It has no common sense
- The more information you give it, the more confused it gets
AI is task—not role—oriented
This is a key concept to understand. AI functions like the proverbial genie in a bottle: a human makes a specific request, the AI completes the task, and then it stops. It doesn’t infer or extrapolate beyond the immediate request. For example, if you ask GenAI to plan a birthday party, it will give you a list of ideas, but it won’t order the cake or send out the invitations. There’s no such thing as an AI event planner or project manager in the way humans understand those roles.
AI has no common sense
Another crucial point is that artificial intelligence lacks common sense. I often introduce executives to the concept of using AI in business as an "infinite interns", but this isn't quite right. Even a young person freshly out of college would know things that AI will not. It's not human, and people who expect it to act human are often disappointed, especially if they let AI do their work for them.
The more information you give it, the more confused it gets
This last limitation is counterintuitive, when looked at from a human perspective. People resolve misunderstandings by communicating more and clarifying needs and what went wrong. For generative AI, especially earlier models, its ability to learn "on the job" is described in the number of "tokens" it can handle. Tokens aren't exactly word count, but it's a good proxy. It has a certain amount of capacity to handle a conversation before completely melting down. This charming meme demonstrates how it often works:

The Bad News
While that may sounds reassuring there are unfortunately a number of jobs that lend themselves very well to replacement by GenAI:
- Fast food drive-thru attendant
- Customer support (both phone and chat)
- Translation of documents
- Categorization and filing of documents
- Document summarization
- Executive assistants
- Non-emergency medical check ups
- Screening candidates for hiring
I want to focus on candidate screening briefly for hiring, as it’s becoming increasingly common. Despite HR slogans like "it’s about people," many job applicants first interact with an AI—or may not interact with a human at all. According to recent data, nearly 99% of Fortune 500 companies are currently using AI to screen job applicants. Video interviewing by AI is so widespread, that there are services to allow candidates to practice and receive AI feedback in advance, so they can prepare!
The trend of AI usage is driven by the power imbalance between employers and candidates, and this pattern will almost certainly continue.
So will I be replaced?
As it stands today, publicly available AI systems aren't ready to take on most human jobs and even if they could, it would likely be too expensive. So where does leave us? For the time being, it's likely that humans and AI will work together; we do things AI can't—for less.
That said, we are at the beginning. Every industry will be transformed, and it remains to be seen who will come out on top. There's no question that our relationship with AI will continue to evolve, but even with the tools we have today, there is plenty to be done. We'd be happy to help.