Newton Institute of Technology Logo
Newton Institute of Technology Building Your Dream Career

Will AI Replace Software Developers? Skills Students Need

Will AI Replace Software Developers?

What students should learn instead of worrying

What happens to programming jobs when AI can write code? Explore the evidence, Kenyan career context, and a tested beginner exercise that shows which skills really matter.

From generated code to reliable software

The valuable skill is understanding whether a solution actually works.

Prepared by Newton Institute of Technology Editorial Team

Updated 11 November 2026

Reading level Beginner / KCSE leaver

Focus AI · Coding · Careers

Quick answer

Will AI replace software developers?

AI is automating some programming tasks and may reduce demand for certain jobs. But available evidence does not establish that all software developers will disappear. AI can generate code and tests, while useful software still needs clear requirements, verification, security and maintenance. Students should learn programming, problem-solving, debugging and responsible AI use—rather than merely copying generated code.

  • Learn one language
  • Understand requirements
  • Test unusual inputs
  • Build complete projects
  • Review AI output

01 / THE TECHNOLOGY

What can AI already do for software developers?

Imagine a shop owner wants a website that displays products and accepts orders. An AI coding assistant can help generate the page, calculate totals, suggest database tables and write initial tests. More advanced coding agents can inspect files, attempt changes and prepare work for review. [2]

AI use is already widespread among respondents to the 2026 Stack Overflow Developer Survey: approximately 83% of respondents to its workplace AI-use question reported using at least one AI tool. Among respondents using coding assistants, 73% said they used them daily. These percentages describe survey respondents, not all developers worldwide. [1]

Task AI assistance What needs checking
Writing code Functions, pages and components Whether the code follows the requirements
Debugging Possible causes and fixes Whether the root problem has been corrected
Testing Suggested tests and scripts Whether important conditions are covered
Database work Tables, queries and relationships Accuracy, access controls and consistency
Maintenance Changes across existing files Whether unrelated features still work

On smaller screens, scroll the table sideways to see all columns.

Writing code is not the same as delivering software.

A school-fee system may calculate KSh 30,000 − KSh 18,000 = KSh 12,000 correctly, yet fail if it records a payment twice. Someone must define the rules and verify those behaviours.

02 / JOB REALITY

Does automating coding mean developers lose their jobs?

Task automation, job redesign and job loss are three different outcomes. A developer may discuss requirements, write code, investigate errors, test changes and deploy a system. If AI makes one activity faster, an employer may deliver more projects, reorganise responsibilities or reduce hiring for some tasks. A successful demo does not prove that an entire occupation has disappeared. [9]

01

Can AI perform the task?

A generated login form may be easy to produce.

02

Is the result reliable?

The form must still handle errors and access controls.

03

Can an employer deploy it?

Security, cost, infrastructure and support matter.

04

What happens to staffing?

A business may grow, reorganise or hire fewer people.

What do employment forecasts say?

The US Bureau of Labor Statistics projects different employment trends for different technology occupations between 2025 and 2035. These are US forecasts, not Kenyan job counts , and they do not isolate AI as the cause. [3]

US occupation Projected change (2025–2035) What it suggests
Software developers +10.2% Broader application-development roles are projected to grow.
Software QA analysts and testers +5.7% Testing work is also projected to grow.
Computer programmers −7.3% A narrower programming occupation is projected to decline.

On smaller screens, scroll the table sideways to see all columns.

Why are students worried about junior roles?

In August 2026, the Stanford Digital Economy Lab reported a roughly 19% relative employment shortfall for US workers aged 22–25 in highly AI-exposed occupations compared with a benchmark based on less-exposed peers. Weaker hiring was an important part of the observed difference. This does not mean 19% of all junior software developer jobs vanished, and the study does not establish that AI alone caused the divergence. [4]

What about productivity?

In selected open-source tasks, a 2025 METR study found experienced developers took 19% longer with the AI tools available at the time. A February 2026 follow-up described signs of greater benefit from newer systems but could not provide a reliable updated estimate. AI usefulness depends on the task, tool and verification effort. [5] [6]

03 / KENYAN CONTEXT

What does AI mean for software developers in Kenya?

A December 2024 ICT-sector study prepared by Genesis Analytics with Kenya’s ICT Authority and other partners estimated around 58,000 people working in software development in Kenya at the time of its assessment. It identified growth opportunities alongside AI-assisted changes and routine-task automation risks. This is an historical research estimate—not a verified November 2026 headcount. [7]

Kenya also launched its National Artificial Intelligence Strategy 2025–2030 in March 2025, reflecting the importance of relevant digital and AI skills. [8]

School systems Accurate records

Marks, access permissions and corrections must be handled correctly.

Retail software Dependable totals

Sales, inventory and refunds must follow clear rules.

Agriculture Useful data

Delivery and payment records must handle missing entries.

These are illustrative software problems, not verified vacancies. International employment statistics cannot be directly applied to Kenya. The better preparation is to learn skills you can demonstrate and to investigate actual employer requirements.

04 / YOUR LEARNING PRIORITIES

Which skills should students learn instead of worrying?

No programming language guarantees protection from automation. Build a foundation that helps you create useful software and adapt when the tools change.

01

Programming fundamentals

Use Python or JavaScript to learn variables, conditions, loops, functions and data structures.

02

Problem definition

Decide the input, rules and expected output before requesting code.

03

Debugging and testing

Reproduce a fault, correct its cause and test important boundaries.

04

Systems understanding

Learn how interfaces, databases, APIs, permissions and backups fit together.

05

Communication and documentation

Explain decisions, record changes with Git and show other people how your software works.

06

Responsible AI use

Ask for explanations, evaluate suggestions, protect private information and verify results.

A simple readiness checklist

If you are a beginner, do not worry about mastering a dozen programming languages. Aim to demonstrate these abilities in one small project.

  • I can explain what my program is supposed to do.
  • I understand the main conditions and calculations.
  • I know how to run my code and reproduce an error.
  • I can test invalid or unusual inputs.
  • I can explain what AI suggested and what I verified.

05 / HANDS-ON ENGINEERING

Practical exercise: find the mistake in this Python program

A fictional school wants to classify examination marks. Valid scores are whole numbers from 0 to 100. A mark of 50 or more means Pass ; a lower valid mark means Fail . Any other input should be rejected.

01

Inspect the code

Look for missing rules and incorrect results.

02

Correct validation

Check the input before deciding Pass or Fail.

03

Run the tests

Verify boundaries and invalid values.

Step 1 — Examine a plausible but incomplete solution

results_checker.py — first draft
def check_result(score):
if score >= 50:
return “Pass”
return “Fail”

For 80 it returns Pass, and for 40 it returns Fail. But 120 also returns Pass, even though the maximum valid mark is 100. Entering the string "80" causes a Python error instead of the required validation message.

Input Original behaviour Required behaviour
80 Pass Pass
40 Fail Fail
120 Pass Invalid score
−10 Fail Invalid score
"80" Raises an error Invalid score

On smaller screens, scroll the table sideways to see all columns.

Step 2 — Validate the input before deciding

results_checker.py — corrected function
def check_result(score):
if type(score) is not int:
return “Invalid score”
if score < 0 or score > 100:
return “Invalid score”
if score >= 50:
return “Pass”
return “Fail”

The first condition rejects values that are not Python integers. The second rejects numbers outside 0–100. The final conditions decide the result. This exercise accepts whole-number marks only; decimal marks would require different requirements.

Step 3 — Test boundary and invalid values

results_checker.py — tests
tests = [
(0, “Fail”), (49, “Fail”),
(50, “Pass”), (100, “Pass”),
(101, “Invalid score”),
(-1, “Invalid score”),
(“80”, “Invalid score”),
(True, “Invalid score”)
]
for score, expected in tests:
assert check_result(score) == expected
print(“All 8 tests passed”)

Expected terminal output: All 8 tests passed

These eight assertions were run successfully in Python. They cover the stated rules, but they are not proof that a much larger school system is free of defects.

Your challenge: add grades A (80–100), B (60–79), C (50–59) and F (0–49). Keep the invalid-input checks and test values 49, 50, 59, 60, 79 and 80. For syntax help, consult the official Python control-flow tutorial . [11]

06 / EVERYDAY PRACTICE

Two more real-world problems beginners can solve

Example A — A small-shop discount calculator

Four products at KSh 250 each total KSh 1,000 . A valid 10% discount subtracts KSh 100, leaving KSh 900 . The real challenge is deciding what happens with a negative quantity, a discount above 100%, or text where a price should be.

01

Define the rules

Write down permitted quantities, prices and discounts.

02

Build the calculation

Compute quantity × price and apply a valid discount.

03

Test the boundaries

Try zero, negative and unusually large values.

04

Review with AI

Ask for a missing test, then verify the suggestion.

Example B — A school rainfall recorder

A science club records 2 mm on Monday, 0 mm on Tuesday, no measurement on Wednesday, and 5 mm on Thursday. The recorded total is 7 mm , and the average across the three measured days is approximately 2.33 mm . Missing information must not automatically be treated as zero rainfall.

Day Rainfall How to interpret
Monday 2 mm Measured
Tuesday 0 mm Measured zero
Wednesday Missing Not measured
Thursday 5 mm Measured

On smaller screens, scroll the table sideways to see all columns.

Try it: store readings in a list, count valid measurements and calculate the measured-day average. If an AI suggestion divides 7 by 4, explain what assumption it made and why that gives the wrong measured-day average.

07 / HOW TO STUDY

Can using AI too much weaken your coding skills?

A controlled study published by Anthropic researchers in January 2026 involved 52 mostly junior software engineers learning an unfamiliar Python library. Participants given AI assistance scored about 17% lower on a later concepts quiz than those who coded without AI. The study was limited in size and scope; it does not establish that AI always reduces learning. The way participants used AI mattered: asking explanatory questions was associated with stronger understanding than delegating most of the work. [10]

01

Attempt independently

Try the task yourself before asking for help.

02

Ask AI for a hint

Request an explanation rather than a full solution.

03

Verify without AI

Rebuild, test and explain what you learned.

A useful learning prompt

“I am learning Python. Review my examination-results function for one possible mistake, explain the reasoning and offer a hint. Do not rewrite the full answer yet.”

After correcting the problem, close the assistant and explain the main logic aloud. If you cannot, practise that concept again. Do not paste confidential records, account credentials or real private student data into AI tools without proper authorisation and controls.

08 / START HERE

Your four-week beginner learning plan

Use this introductory schedule to explore programming and build evidence of progress. It is not a complete professional training programme.

Week Learning focus Proof of progress
1 Variables, conditions, functions Build and explain a simple calculator
2 Validation, debugging, tests Complete the school-results checker
3 AI-assisted review Document one accepted and one rejected AI suggestion
4 Build and present a project Demonstrate the shop or rainfall program with tests

On smaller screens, scroll the table sideways to see all columns.

Practise approximately 45–60 minutes on most days, adjusting the schedule to your available time. Spend some time writing independently, some reviewing documentation or AI suggestions, and some running tests.

Five questions to ask yourself after four weeks

What problem does my program solve? What information does it receive? How does its logic work? Which unusual inputs have I tested? What would I improve next?

09 / CAREER DECISION

Is software development still a sensible career choice?

It can be a rewarding field for students who enjoy creating things, investigating problems and learning new tools. But no programme, programming language or AI tool guarantees employment. Job availability differs by location, employer, experience and specialisation.

If you are considering software development after KCSE, inspect course content, teaching time, project opportunities, attachment arrangements and admission requirements. Try the four-week challenge before deciding which deeper pathway suits you.

Show proof—not just screenshots

A beginner portfolio is more convincing when you can explain your decisions and demonstrate the software working.

  • State the problem and its rules.
  • Demonstrate a working feature.
  • Show tests for unusual inputs.
  • Explain one mistake you fixed.
  • Describe what AI assisted with.

For more about junior roles and project presentation, see NIT’s junior full-stack developer guide and coding portfolio guide .

10 / QUICK QUESTIONS

Frequently asked questions

Will AI completely replace software developers by 2030?

No reliable evidence establishes that all software developers will disappear by 2030. Automation may expand, but outcomes will vary across tasks, companies and countries.

Should I still learn programming if AI can write code?

Yes, if your goal is to build or understand software. Programming knowledge helps you recognise mistakes, evaluate requirements and test results.

Are junior developer jobs more at risk?

Some entry-level tasks can be automated. Research shows employment pressure for some young US workers in AI-exposed occupations, but it does not establish that every junior developer position will disappear.

Can AI build an entire application?

Some agents can attempt substantial application development. Whether the result is dependable requires testing, security review and context-specific checks.

Should I begin with Python or JavaScript?

Both are appropriate. Python works well for introductory logic and automation; JavaScript is particularly useful for interactive web applications. Learn one well before expanding.

Do I need to study artificial intelligence first?

No. Begin with programming fundamentals and progressively learn how AI assistants fit into software development.

Will AI lower software developer salaries?

No single salary forecast applies to everyone. Compensation depends on demand, location, experience, business decisions and specialisation.

Can I start software development without a university degree?

You can begin learning through technical courses and guided projects, though job qualification requirements differ. Check the positions you hope to pursue.

The bottom line

Learn to understand, test and improve software.

AI is changing development work, and some career routes may become harder. But it is still useful to understand how programs work, what users need, and how to discover when a result is wrong. Start with one language, solve one small problem, and be able to explain what you built.

The goal is not simply to type code faster than AI. It is to know what needs building, how to test it and why the solution works.

Study at NIT · Migori Town

Start building practical software-development skills.

Newton Institute of Technology offers Full-Stack Software Development training covering frontend development, backend programming, databases, APIs, testing, deployment and practical projects. Develop the foundations needed to understand and evaluate the software you build.

Institutional training 8 months

Industrial attachment At least 3 months

Published course fee KSh 90,000

Published programme information is provided in NIT’s official course guide; confirm the latest admissions arrangements with the institution. [12]

Sources and references

  1. Stack Overflow (2026). AI tool use and frequency among survey respondents. Stack Overflow Developer Survey 2026: AI data .
  2. GitHub Docs. Official coding-agent capabilities and review workflows. GitHub Docs: Coding agents .
  3. US Bureau of Labor Statistics (2026). Occupational Employment Projections: Computers and Information Technology. BLS 2025–2035 ICT projections .
  4. Brynjolfsson, E., Chandar, B. and Chen, R. (2026). Young workers and AI-exposed employment trends. Stanford Digital Economy Lab, August 2026 .
  5. METR (2025). Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity. METR 2025 productivity experiment .
  6. METR (2026). We Are Changing Our Developer Productivity Experiment Design. METR February 2026 follow-up .
  7. Genesis Analytics (2024). Unlocking Current and Future Employment for ICT Professionals in Kenya; report developed with ICT Authority and partners. Kenyan ICT employment research (2024) .
  8. Ministry of Information, Communications and the Digital Economy (2025). Kenya’s Artificial Intelligence Strategy launched. Kenya AI Strategy 2025–2030 .
  9. International Labour Organization (2025). The Future of Work: How Will Junior Programmers Be Affected? ILO on AI exposure and junior programmers .
  10. Anthropic Research (2026). How AI Assistance Impacts the Formation of Coding Skills. Anthropic study of AI-assisted coding skills .
  11. Python Software Foundation. The Python Tutorial — More Control Flow Tools. Python Tutorial: control flow .
  12. Newton Institute of Technology (updated 2026). Full-Stack Software Development Course. NIT full-stack programme .