Future of ICT careers
Which ICT Jobs Will AI Change Most—and Which Skills Will Become More Valuable?
Jobs · Skills · Practical projects
See how AI is changing software, data, support and cybersecurity work, what professionals still need to do, and how students can practise the skills that matter.
The modern ICT workflow
Use AI. Check the result. Take responsibility.Quick answer
Which ICT jobs will AI change most?
Routine coding, basic helpdesk responses, repetitive testing, standard data reporting and data labelling are especially exposed to AI-driven task change. Software development, data analysis, cybersecurity, cloud operations and networking are also evolving—but that is not the same as whole careers disappearing. Skills that matter include technical foundations, problem-solving, testing, AI literacy, cybersecurity, data reasoning and communication. [1] [2]
- Compare ICT jobs
- Understand junior-role changes
- Practise with real scenarios
1. Which ICT jobs will AI change most?
The biggest immediate changes tend to affect repetitive, information-based tasks. Here is how the work is likely to shift across ten ICT career paths. This is a qualitative task comparison, not a ranking of how many jobs will be lost.
| ICT occupation | Tasks AI may assist or automate | Human skills to strengthen |
|---|---|---|
| Data annotation and labelling | Repetitive classification and standard labels | Data quality, hard-case review, subject knowledge |
| IT support technician | Routine helpdesk responses and ticket sorting | Diagnostic networking, user support, escalation |
| Software developer | Routine functions, boilerplate code, documentation | Logic, system design, debugging, secure databases |
| Web / mobile developer | Templates, interface components and page drafts | Usability, accessibility, integration, testing |
| Software tester | Test drafts and repetitive checks | Exploratory tests, risk thinking, error analysis |
| Data analyst | Formulas, charts, report drafts and queries | SQL, statistics, data validation, interpretation |
| Cybersecurity analyst | Alert grouping, logs and investigation summaries | Identity security, incident response, evidence |
| Network / systems administrator | Configuration suggestions, monitoring and scripts | Access controls, safe changes, recovery |
| Cloud / DevOps engineer | Deployment scripts and infrastructure drafts | Reliability, monitoring, security and rollback |
| AI / machine-learning engineer | Prototype code and experiment assistance | Data engineering, model evaluation, monitoring |
On a small screen, swipe horizontally to read the complete table.
2. Why a changing task is not a disappearing job
Consider a school ICT technician. The technician answers standard questions, installs software, troubleshoots networks and maintains computers. An AI assistant may answer a familiar password question, but a laboratory-wide network failure still requires physical checks, diagnosis and approved corrective action.
Automation
The tool performs a task. A helpdesk assistant classifies a ticket or sends standard reset instructions.
Augmentation
The tool assists a professional. AI suggests a SQL query; a data analyst checks it against the actual records.
Transformation
The job mix changes. Less repeated drafting may mean more investigating, checking and explaining.
Ask four questions about any task: Is it repetitive? Is the information digital and authorised for use? Can an answer be verified? What happens if the result is wrong? The more serious the possible consequences, the stronger oversight should be. [1] [3]
3. Software development: faster drafts, but reliability still matters
AI coding tools can write simple functions, explain errors, suggest tests and draft documentation. They do not independently establish that a complex application meets its business requirements. Developers still need programming logic, databases, debugging, system design, security and testing.
Example: A school fee calculator with duplicate payments
A learner owes KSh 10,000. Confirmed payments total KSh 7,000, so the correct balance is KSh 3,000. But if a KSh 3,000 payment is accidentally entered twice, a program that blindly sums records may report zero. The underlying data—not merely the arithmetic—must be examined.
- Define the expected balance. Start with verified fees and confirmed payment entries.
- Check the input records. Investigate duplicates using transaction references; never silently delete real financial records.
- Test unusual cases. Try missing records, reversals and erroneous values in a fictional laboratory.
- Verify the result. Compare the program output with an independent calculation before accepting it.
Speed is not guaranteed. In a 2025 controlled study of experienced developers working on familiar open-source projects, METR found tasks took 19% longer when AI assistance was allowed. In a February 2026 update, METR noted that newer tools might improve productivity, but methodological selection effects prevented a reliable estimate. These findings are specific to those studies—not to every AI tool or programmer. [4] [5]
4. Web and mobile development: a polished screen can still fail
AI can create a website layout or a mobile form. A developer must still evaluate accessibility, usability, data handling, application logic and testing.
Example: Two clients book the same salon appointment
A student creates a booking page that looks correct. Two clients submit the same appointment time and both receive confirmations. The interface works, but the system has a booking-conflict defect. The developer must understand how bookings are stored and prevent conflicting reservations.
- Reproduce the conflict. Submit two requests for the same slot in a controlled test.
- Inspect the rule. Determine where availability is checked and when the reservation becomes final.
- Correct the application. Use an appropriate database and application-level strategy to prevent double bookings.
- Retest. Only one request should succeed for a single available slot; the other should receive a clear message.
Skills to practise: HTML, CSS, JavaScript, mobile interfaces, databases, APIs, accessibility and integrated testing.
5. Software testing: AI can draft tests, but who decides the risks?
Automated tools can generate many tests. Yet even hundreds of passing tests do not show that every important scenario was checked. The booking application above could pass a normal single-user test and still fail when two people act simultaneously.
Normal case
A customer enters valid data and submits successfully.
Unusual case
Two users book the same appointment at nearly the same time.
Failure case
The database is unavailable, so the system must not pretend the booking was saved.
A tester must choose meaningful cases, question assumptions and investigate results. Skills to practise: exploratory testing, test design, debugging and risk-based verification. [3]
6. IT support: simple answers are easier than real troubleshooting
AI can draft instructions for printers, passwords and software errors. A technician still needs to identify which explanation is supported by the observations.
Example: One computer loses internet access
- Determine the scope. Check whether nearby computers can connect; avoid disrupting everyone.
- Inspect the device. Check its cable or Wi-Fi connection and approved network settings.
- Run appropriate tests. Examine IP configuration and connectivity or name-resolution behaviour.
- Compare explanations. Use AI to suggest possibilities, then compare each with your evidence.
- Correct and document. Apply an authorised correction and verify the connection again.
Skills to practise: hardware, Windows or Linux, IP networking, DNS, diagnosis and user communication.
7. Data analytics: AI can make charts, not incomplete data trustworthy
A report may be beautifully formatted and still misleading. Suppose AI compares sales at three supermarket branches and shows that Branch B performed worst. Before recommending action, an analyst discovers that Branch B is missing an entire day of transactions. The summary describes the supplied spreadsheet, not necessarily the business reality.
- Understand the question. Which period and which branches are being compared?
- Check source records. Look for missing days, duplicates and incorrect values.
- Verify the analysis. Recalculate totals and inspect the AI-generated chart.
- Explain limitations. Report what the evidence supports and what still needs investigation.
Skills to practise: Excel, SQL, Python, statistics, data quality and business explanation.
What about data annotation?
Data annotation means labelling examples, such as identifying objects in images or categorising text. Repetitive labels may be automated, while difficult cases and quality evaluation can require human input. A December 2024 Kenya-focused study anticipated both short-term opportunity and greater longer-term automation exposure for data annotation. These are different measures, not a contradiction. [8]
8. Cybersecurity: AI flags problems; people investigate evidence
Security AI may summarise logs, group alerts and highlight unusual sign-ins. An alert does not, by itself, prove an attack.
Example: Five failed administrator logins
A school system records five unsuccessful attempts to access an administrator account. An AI tool calls this suspicious. A legitimate user may have forgotten a password; an unauthorised person may also be attempting access. A trained analyst checks records, follows incident procedures and escalates when appropriate.
Observe
What do the logs actually say? Record the time, account and result.
Investigate
Which further evidence is authorised and relevant? What are the alternative explanations?
Respond
Follow approved incident rules and document only conclusions supported by evidence.
AI agents that can use tools and access systems introduce additional concerns about permissions and untrusted instructions. The OWASP agentic-AI security guidance covers these risks. [11]
Skills to practise: networks, operating systems, identity security, log analysis, incident response and safe AI system permissions. Only use systems you own or are authorised to test.
9. Network, cloud and AI engineering: automation needs oversight
AI can draft configuration scripts, deployment steps and monitoring summaries. However, an incorrect infrastructure change can interrupt services. When a school registration site goes down, a professional must identify the fault, examine the suggested fix, obtain required approval, prepare recovery steps and verify the outcome.
Network and system administration
Learn IP networking, access control, Linux, safe configuration and backups.
Cloud and DevOps
Learn deployments, monitoring, reliability, rollbacks and infrastructure security.
AI / machine-learning engineering
Learn Python, data preparation, statistics, model evaluation and production monitoring.
AI engineering is also affected by AI. Tools may accelerate prototype code, but incorrect data and inappropriate model evaluation can still produce unreliable results.
10. Will AI make it harder to get a first ICT job?
Some junior roles include routine tasks that AI tools can now assist with. That raises a serious question: will employers offer fewer opportunities for newcomers to learn those tasks?
A Stanford Digital Economy Lab working paper revised in August 2026 examined a US payroll-data sample through June 2026. Employment for workers aged 22–25 in AI-exposed occupations was 19% below a comparison trend based on less-exposed peers. The researchers also found no evidence of widespread economy-wide displacement; the difference appeared more in hiring than separations. Their findings are descriptive, not proof that AI caused all the observed changes. They are US data, not Kenyan employment statistics. [6]
An additional 2026 international jobs analysis describes skills requirements changing quickly in AI-exposed work, including some junior positions. [7]
11. Which ICT skills become more valuable?
The World Economic Forum identifies AI and big data, networks and cybersecurity, and technological literacy as fast-growing skill areas through 2030. Analytical thinking, adaptability, creativity and ongoing learning also matter. [2]
| Skill | Why it matters | How to demonstrate it |
|---|---|---|
| Technical foundations | Understand code, systems, databases or networks | A working program or documented lab |
| Problem-solving | Investigate causes instead of guessing | Fault log with observations and tests |
| AI literacy | Use AI appropriately and question its output | Comparison of your solution and AI suggestion |
| Testing and verification | Discover incorrect behaviour | Test cases and fixed errors |
| Data reasoning | Check records and interpret results | Cleaned spreadsheet and short explanation |
| Cybersecurity and privacy | Protect accounts and information | Permissions and security checklist |
| Communication | Explain evidence and decisions | Problem → evidence → action → result report |
| Adaptability | Learn new tools without losing fundamentals | Improved version of an existing project |
On a small screen, swipe horizontally to read the complete table.
A student who can prompt AI to write SQL but cannot read the query may miss an error. A student who understands SQL can request assistance, inspect the logic, test it and explain the result. Prompting helps; understanding and verification make it dependable.
12. A simple workflow for using AI responsibly
The most transferable habit is not a particular prompt. It is a repeatable method for getting dependable outcomes.
- Understand. Write down the problem and the result you expect.
- Ask. Use an authorised AI tool for explanations, draft code or possible causes.
- Check. Test the answer using source data, real observations or suitable documentation.
- Correct. Find and fix errors rather than repeatedly accepting new guesses.
- Explain. Record what AI helped with, what you verified and why you accepted the solution.
13. Three practical AI-ready ICT exercises
These exercises use invented data. They are designed to develop verification, troubleshooting and test-design habits without exposing real systems.
Exercise A — Detect an incorrect school-fee balance
A student owes KSh 10,000. Enter the following fictional payment records:
| Payment reference | Amount (KSh) |
|---|---|
| TEST101 | 2,000 |
| TEST102 | 1,500 |
| TEST103 | 3,000 |
| TEST102 | 1,500 |
| TEST104 | 1,000 |
- Prepare the worksheet. Record the fictional payments exactly as shown in the table.
- Calculate the initial totals. Sum all entries and subtract that amount from KSh 10,000.
- Look for duplicate references. Use a spreadsheet function or ask AI how to highlight repeated values.
- Investigate the duplicate. For this classroom scenario, assume the second TEST102 entry is confirmed to be an accidental copy.
- Calculate the corrected result. Exclude only the confirmed accidental copy and compare the result with an independent calculation.
- Explain the lesson. A duplicated reference is a signal for investigation—not permission to delete real transactions.
Exercise B — Test an AI-generated registration form
Build a simple fictional school-event registration form. Ask AI for suggestions if helpful, but test a valid submission, missing name, no activity selected, double submission, simulated server failure and a narrow mobile screen. Record results in a test log and correct at least one weakness.
Expected output
A working form plus a test checklist covering normal and failure cases.
Skills demonstrated
HTML, form logic, usability, testing and explaining a defect.
Exercise C — Measure whether AI actually helps
Choose two similar small exercises such as spreadsheet formulas or short programming functions. Complete one independently and the other with AI assistance. Record time to first draft, verification time, errors, corrections and final correctness. The comparison is a learning exercise—not a controlled scientific study.
14. What does Kenya-specific research say?
The December 2024 report Unlocking Current and Future Employment for ICT Professionals in Kenya, authored by Genesis Analytics in collaboration with Kenyan and development partners, separated employment potential from exposure to automation. Its executive-summary assessments include the following: [8]
| ICT job family | Report assessment |
|---|---|
| Software development | AI augmentation expected |
| Data science and analytics | AI augmentation expected |
| Data annotation and labelling | Significant automation exposure over time |
| Cybersecurity | AI augmentation expected |
| AI governance and ethics | Limited exposure |
| Data-centre technicians | AI augmentation expected |
| Networking and architecture | AI augmentation expected |
| IT operations and support | AI augmentation expected |
These are 2024 research assessments, not observed job losses in 2026. They are especially useful because they distinguish repetitive data labelling from occupations in which AI augmentation was expected.
Kenya’s Ministry of ICT identifies programming, networking, cybersecurity, AI, big data and mobile development among advanced digital skill areas. The National AI Strategy 2025–2030 also addresses infrastructure, data governance, research and innovation. [9] [10]
Apply your skills to realistic local scenarios: a school fee register, a shop sales report, a secure customer database, a water-production inventory system or a network troubleshooting case. These are practice opportunities, not promises of specific vacancies.
For a separate overview of market demand, read NIT’s Most In-Demand IT Fields in Kenya guide.
15. How do you choose which ICT field to learn?
Select a field by the problems you enjoy solving and the foundations you are willing to practise—not by claims that a career is permanently safe from AI.
| If you enjoy… | Explore… | First small project |
|---|---|---|
| Building websites and apps | Full-stack or mobile development | A booking form or task-list app |
| Working with numbers | Data science and analytics | A verified shop-sales dashboard |
| Protecting computer systems | Cybersecurity | A permitted security-lab report |
| Investigating faults | IT support and networking | A documented connection diagnosis |
| Exploring intelligent systems | Artificial intelligence | A simple Python/data experiment |
On a small screen, swipe horizontally to read the complete table.
When choosing training, check programme content, available practical work, equipment, eligibility requirements and opportunities for supervised experience.
16. A four-week practical learning challenge
This is an introductory study plan, not a replacement for a full ICT course or industrial attachment.
- Week 1 — Learn. Choose one ICT task, learn its foundations and attempt a small independent solution.
- Week 2 — Compare. Use AI to suggest an alternative; note what is different and what you understand.
- Week 3 — Verify. Test ordinary and unusual cases, investigate one failure and correct it.
- Week 4 — Demonstrate. Show the project and explain your contribution, AI assistance, tests and outcome.
A project report can use five headings: Problem, Tools, My contribution, Verification, Result. A small project you can explain is stronger beginner evidence than a large generated application you cannot understand.
17. Frequently asked questions
Which ICT jobs will AI change the most?
Jobs containing repetitive digital tasks—such as basic helpdesk responses, routine coding, some testing, reporting and data labelling—may see substantial task change. This does not state how many employees will be replaced.
Will AI replace junior programmers?
AI changes some tasks commonly assigned to junior developers. US research has found hiring pressure for young workers in some AI-exposed occupations, but does not show that every junior programming role will disappear. Learning debugging, testing and fundamentals remains valuable.
Is data analysis still useful when AI creates reports?
Yes. Analysts still need to check data quality, choose appropriate calculations, interpret context and explain what evidence does or does not show.
Is cybersecurity safe from AI?
No job is guaranteed to remain unchanged. AI assists some cybersecurity tasks while creating new issues around identity, permissions and securing connected AI systems.
Must every ICT student become an AI engineer?
No. AI literacy can support software, networking, cybersecurity, support and data roles without requiring every learner to specialise in machine learning.
What matters more than knowing prompts?
Understanding the task, checking results and correcting errors. Prompting is useful; independent verification builds dependable skill.
Is ICT still worth studying in Kenya?
ICT skills remain relevant to many Kenyan sectors. Assess specific programmes and local opportunities, practise with realistic projects and avoid assuming a course alone guarantees employment.
Final advice: learn the work, not only the AI tool
AI can produce code, reports and suggested solutions quickly. But the speed of a first answer does not establish that the result is accurate, useful or safe. Learn the fundamentals of your chosen field, practise with real-world problems, use AI when it helps, test what it produces and explain what you have verified.
Your next step
Build practical ICT skills at Newton Institute of Technology
Prepare for technology work that increasingly uses AI. Explore training in Full-Stack Software Development, Cybersecurity, Data Science & Analytics, Artificial Intelligence and Mobile App Development at Newton Institute of Technology in Migori Town, Kenya.
Sources and references
- [1] International Labour Organization (2025). Generative AI and Jobs: A Refined Global Index of Occupational Exposure. Task-level exposure and job transformation.
- [2] World Economic Forum (2025). Future of Jobs Report 2025, Jobs Outlook and Skills Outlook. Forecast occupational and skill changes.
- [3] Skills England (2026). Sector Skills Needs Assessment: Digital and Technologies. Technical-task changes, verification and oversight.
- [4] METR (10 July 2025). Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity.
- [5] METR (24 February 2026). We Are Changing Our Developer Productivity Experiment Design. Newer-tool evidence and methodological limitations.
- [6] Stanford Digital Economy Lab (revised 12 August 2026). Canaries in the Coal Mine? Six Facts About the Recent Employment Effects of Artificial Intelligence. US payroll-data study.
- [7] PwC (15 June 2026). 2026 Global AI Jobs Barometer. AI-exposed occupations and changing skills.
- [8] Genesis Analytics (December 2024). Unlocking Current and Future Employment for ICT Professionals in Kenya, executive summary p. 4.
- [9] Kenya Ministry of Information, Communications and the Digital Economy. Digital Skills, official programme priorities.
- [10] Kenya Ministry of ICT (2025). Kenya National Artificial Intelligence Strategy 2025–2030, launch notice.
- [11] OWASP (2026 edition). OWASP Top 10 for Agentic Applications, security guidance.
Research estimates and projections are not job-loss guarantees. US findings are not Kenyan employment statistics. Practical examples are fictional exercises; NIT programme information should be checked on the official admissions pages.