AI has made software creation faster and more accessible, but speed is not the same as quality.
The competitive advantage is shifting from who can write code to who can define the problem, guide the build, test the output, manage risk and keep the system maintainable.
Euphoria Telecom technology director Nic Laschinger says there are five key factors to consider before using AI to develop software.
AI-built software is already here
The 2025 South African GenAI Roadmap found that GenAI adoption among large local enterprises had climbed to 67% in 2025, up from 45% in 2024.
Code generation is also gaining ground, with 36% of respondents already using GenAI for this purpose, up from 24% in 2024. Yet only 14% had a clearly defined strategy for integrating GenAI into business operations. In other words, most local businesses are still working out how to manage the risks that come with AI.
Software expertise is becoming more strategic
Gartner predicts that by 2028, 90% of enterprise software engineers will use AI code assistants, up from less than 14% in early 2024. The developer role will shift from implementation to orchestration, with more focus on problem-solving, system design and making sure AI tools produce quality outcomes.
This changes the value of software expertise. Teams will spend less time manually producing every line of code and more time asking better questions. What exactly are we building? Who will use it? What data will it access? How will it connect to existing systems? What happens if it fails? Can it scale? Can it be maintained? Can it be secured?
Productivity gains are real
Stack Overflow’s 2025 Developer Survey reported that 52% of developers believe AI tools or agents have had a positive effect on their productivity.
A McKinsey study found that developers using generative AI tools could complete some routine coding tasks up to twice as fast.
Documenting code could be done in half the time, writing new code in nearly half the time, and refactoring existing code in about two-thirds of the time.
AI can slow teams down when trust is weak
Speed gains have not yet created confidence. Stack Overflow found that 46% of developers actively distrust the accuracy of AI tools. A 2025 METR trial found that experienced open-source developers working on familiar repositories took 19% longer with AI tools than without them, a warning against assuming AI always speeds up complex work.
The real risk is hidden technical debt
Security is one of the clearest risks. Veracode’s 2025 GenAI Code Security Report found that 45% of AI-generated code samples failed security tests and introduced OWASP Top 10 vulnerabilities.
AI-generated code should therefore be treated as untrusted until it has been reviewed, tested and secured.
The same applies to maintainability. AI may produce code that works in a demo but creates problems later if it is repetitive, poorly integrated or hard to understand. The cost shows up when the business needs to add features, connect systems, protect customer data or recover from a security issue.
The value of AI ultimately depends on how well teams instruct the tool, test its output, manage risk and build it into daily delivery. Strong teams can use AI to take pressure off routine work and move faster without losing control of quality. Weaker teams are more likely to face extra checking, avoidable rework and problems that only surface later.
The test for business leaders is discipline. AI makes it easier to start building, but it also makes it easier to skip the hard thinking that delivers good software. To benefit, your business needs to treat AI as part of a controlled delivery process that includes clear ownership, proper review and a realistic view of long-term risk.