The rise of the AI-enabled amateur
This shift isn't just anecdotal anymore. There is now measurable evidence that people outside traditional technical roles are using AI to cross into work that previously required specialised skills.
OpenAI's June 2026 analysis of Codex usage provides a striking example.
Between August 2025 and early June 2026, the number of individual non-developer Codex users grew 137 times, while non-developer users within organisations grew 189 times. OpenAI also found that employees in non-technical departments were increasingly using Codex for tasks such as automation, data transformation, debugging, structured analysis and other technical work.
The important point isn't that 137 times more people suddenly became software developers.
They didn't.
Rather, people who weren't developers started using a tool originally built around coding to perform work that previously sat outside their job descriptions.
OpenAI's research describes this as an expansion of what an individual worker can do. In business functions, for example, more than one-quarter of the work performed with Codex involved engineering or coding tasks.
That is exactly where the idea of the AI-enabled amateur becomes interesting.
Someone may not be a programmer, but they can now build a prototype.
Someone may not be a data analyst, but they can interrogate a dataset.
Someone may not be an automation specialist, but they can build an automated workflow.
Someone may not be a software engineer, but they can describe what they need and have AI help turn that description into working software.
The person hasn't necessarily acquired the conventional qualification.
But the boundary of what they can actually do has expanded.
And this is happening beyond software.
In 2023, Kenya Solomon was working as a security guard at Kane Realty in Raleigh, North Carolina. He wasn't a software engineer or a technical specialist. But during an AI training programme, he used ChatGPT to tackle a long-standing problem involving the company's fire-alarm systems. By combining a 300-page technical manual with the knowledge of experienced maintenance workers, Solomon built a working AI-powered tool in days. The solution impressed company leadership and eventually helped him move into a newly created Technical Project Manager role
Similarly, Colin Knudsen, co-founder of fleet-software company Proaction, isn't a software engineer. He has used AI coding tools to turn customer requirements into functioning software demonstrations, allowing him to bridge the gap between understanding a customer's problem and technically building a solution.
These examples point towards something much bigger than AI simply making existing professionals faster.
AI is allowing people to cross skill boundaries.
- A marketer can move closer to coding.
- A business owner can move closer to software development.
- A designer can move closer to product development.
- An accountant can move closer to automation.
- A researcher can move closer to data analysis.
The traditional job description starts to become less restrictive.
From "I can't do this" to "Let me try"
For decades, one of the biggest barriers to acquiring a new skill was the first step.
- You needed to know where to start.
- You needed to understand the terminology.
- You needed to find the right tutorial.
- You needed someone to explain what you were doing wrong.
And, often, you needed to be comfortable being bad at something for a very long time.
AI changes that experience.
You can start with:
> "I don't know anything about this. Explain it to me."
Then:
> "Show me an example."
Then:
> "Let me try."
Then:
> "Here is what I made. Tell me what's wrong."
Then:
> "Help me improve it."
That is remarkably close to having a tutor, assistant and junior consultant available simultaneously.
But there is an important warning.
Access to expertise is not the same as possessing expertise.
AI can help someone produce a sophisticated result without necessarily giving them the underlying understanding required to judge whether that result is actually good.
That distinction may become one of the most important issues in the AI era.
The future may therefore belong neither to people who know everything nor to people who know nothing.
It may belong to people who know enough to direct AI, question AI and recognise when AI is wrong.
AI is not eliminating the value of skill.
It is changing the economics of acquiring and applying it.
And perhaps that is what "democratising expertise" really means.
Not giving everyone a PhD.
Giving everyone a chance to start acting like they have access to one.
