AI Can't Save Software Engineering
12 hours ago
- #AI
- #Software Engineering
- #Startups
- AI for Software Engineering (AI4SE) is seen as a potential startup opportunity, but the author believes it's a flawed approach.
- Software engineering is fundamentally about managing people, incentives, and responsibilities, not just code and tools.
- AI can't solve management issues like conflicting KPIs or organizational structure, which are often the root of software problems.
- The value of AI tools in software engineering is hard to measure, making it difficult to justify their cost to stakeholders.
- AI4SE tools often sell 'supplements' (nice-to-haves) rather than 'painkillers' (must-haves), which is problematic in a cost-cutting environment.
- The business model of experts selling to experts (engineers) is challenging and doesn't scale well.
- Small teams should avoid AI startups focused on optimizing software engineering, as they require changing company culture and power dynamics.
- Exceptions like Cursor succeed by focusing on individual developers' immediate needs, avoiding management and ROI pitfalls.