What AI is genuinely good at in software
- Writing routine code fast: forms, standard screens, API calls, boilerplate.
- Explaining and searching code: "where is this calculated?" in a large codebase.
- Drafting tests and documentation.
- Prototypes: a working demo in hours instead of weeks.
- Translating between languages and frameworks.
What it still can't do on its own
- Decide what to build. Understanding the business, the users and the trade-offs is the hardest part of software, and it's human work.
- Own the consequences. Someone must be accountable for security, data protection and uptime.
- Design systems that last: data models, architecture and integrations that stay sound for years.
- Judge its own output reliably. AI code can look right and be subtly wrong; review and testing remain essential.
- Talk to your customers, stakeholders and regulators.
Our view: AI is replacing typing, not thinking. Developers who use AI well are much more productive. Teams that skip review and testing ship problems faster.
How the developer job is changing
| Less time on | More time on |
|---|---|
| Writing boilerplate | Specifying clearly what should be built |
| Looking up syntax | Reviewing and testing AI-generated code |
| Repetitive changes | Architecture, data design and integrations |
| Manual documentation | Security, performance and reliability |
What this means if you're hiring
- Expect more output per developer, so a smaller, stronger team may do what a larger one did before.
- Hire for judgement, not just speed: ask candidates or agencies how they review AI-generated code, test it and handle security.
- Don't remove senior oversight. Junior-only teams with AI tools can produce a lot of fragile code quickly.
- Pay for outcomes: fixed-price projects with clear deliverables, or hourly work with visible progress.
- Ask about AI policy: which tools they use, and how your code and data are protected.
Will it be cheaper to build software?
Building the first version is getting faster and cheaper. But the share of the cost that goes into understanding the business, integration, security, testing and maintenance hasn't gone away; if anything it matters more, because more software is being created. The best value comes from teams that use AI to move fast and still do the engineering properly.
A note on global hiring
AI tools work the same everywhere, which makes the location of your team matter less than its skill and communication. That's one reason more companies in the US and Europe are working with remote teams in places like Kenya, where they get strong engineers, overlapping working hours and lower costs. See outsourcing software development to Africa.
AI won't replace developers who understand problems, design systems and take responsibility for what they ship. It will replace slow ways of working. When you hire, look for a team that uses AI to deliver faster and can explain how they keep quality high. Our team in Nairobi builds web, mobile, AI and custom software for clients worldwide, using AI tools throughout with human review on every change.
See hire developers in Kenya or read vibe coding vs hiring a developer.
