GitHub Copilot, Claude, and similar AI coding tools can write substantial portions of code from natural language descriptions. Here's what that means for your career and what to do about it.
AI is not eliminating programming; it is changing what programmers are paid to do. The volume of code a programmer can produce is rising.
TASK LEVEL RISK
Most of the work stays human. AI assists at the edges.
AI is handling specific tasks. The core role is intact but shifting.
AI is automating significant portions of the work. Adaptation is essential.
Higher risk
routine code implementation from clear specifications, boilerplate and template code generation, standard debugging of common errors, documentation writing, basic unit test generation
Lower risk
requirements analysis and software design, code review and quality assurance, system architecture, security review, debugging complex and novel failures, technical leadership and mentorship
Programmers provide the requirements understanding, architectural judgment, code review, debugging expertise, and accountability for software quality that AI code generation cannot self-assess. The translation between what users need and what software should do, and the judgment to evaluate whether AI-generated code achieves it, are human responsibilities.
WHAT YOU SHOULD DO
Skills to build for the AI era
New skills - Adapt to the AI landscape
New skills - Adapt to the AI landscape
Using AI coding assistants effectively, including prompt engineering for code generation and iterative refinement to produce correct and maintainable code.
Critically evaluating AI-generated code for correctness, security vulnerabilities, maintainability, and alignment with requirements before deploying it.
Crafting precise prompts that produce useful code, understanding AI coding tool limitations, and iterating effectively to achieve the desired implementation.
Timeless skills - What AI can't replicate
Timeless skills - What AI can't replicate
Translating what users need into technical requirements and sound software designs is the most important human contribution before any code generation.
Diagnosing complex failures from system-level interactions rather than isolated functions requires the judgment of an experienced programmer.
Designing the structure of software systems that can grow, be maintained, and meet non-functional requirements requires engineering judgment.
THE FULL PICTURE
What AI can do, what it can't, and where the career is headed
What AI can already do
- Generate functional code from natural language descriptions or specification prompts
- Complete code inline and suggest refactoring improvements
- Write unit tests and generate documentation from existing code
- Identify common bugs, security vulnerabilities, and code style issues
What AI can't do
- Understand what a user or business actually needs and translate it into a sound software design.
- Review AI-generated code for correctness, security vulnerabilities, and maintainability.
- Debug novel and complex failures that emerge from system interactions rather than individual functions.
- Take accountability for software that runs in production and affects users.
The profession is being restructured rather than eliminated.
Do you have the right strengths for this career?
Our test measures your personality and strengths — and shows how you match with 1600+ careers.
Job outlook
BLS projects a 10 percent decline in computer programmer employment from 2024 to 2034 as AI tools enable fewer programmers to produce the same code output. Median annual wages were $99,790 in May 2024. Software developer roles, which require broader design and architecture skills, are growing and partially absorbing programmers who upskill.