Building on the 2023 list of essential “-ilities”, the landscape of software engineering in 2025 demands a refreshed perspective with artificial intelligence (AI) at the center. As AI has moved from niche add-on to foundational pillar in software systems, both risks and opportunities have multiplied.
Why Refresh the “-ilities” for AI?
While the "-ilities" of good system design like stability, scalability, and the rest still matter (maybe more than ever!), widespread AI adoption intensifies the need for transparency, ethical design, and trust. AI isn’t just super-charging productivity and automation, it can also introduce “black box” outputs, new modes of failure, and potentially broad impacts on users, society, and organizations.
Three New Critical “-ilities” for the AI Age
1. Accountability
AI decisions must always be attributable and reviewable. Who “owns” an automated outcome? Who fixes it if something goes wrong? Modern software must keep a log for all AI-driven actions and the reasoning behind those actions, not only for compliance, but to maintain trust.
2. Auditability
AI’s “black box” nature intensifies the need for full traceability. From data lineage and model versioning to explainable predictions and error logs, robust auditability is now a core requirement—especially for regulated domains. Good software logs what the machine and the human using it did, and now we have something new to log -- the reasoning in a sometimes non-deterministic system.
3. Ethicality
Fairness, bias mitigation, and responsible outcomes are now mainstream software requirements, not afterthoughts. Proactively design systems to detect and correct for bias, adhere to ethical standards, and comply with evolving global AI regulations.
The Rest Still Matter… but Look Different in 2025
Explainability (XAI): Goes mainstream—no longer just a “nice-to-have” for data science. End users and regulators expect understandable AI-powered outcomes.
Testability: Encompasses adversarial, robustness, and bias testing, not just classic QA.
Maintainability and Extensibility: Now extend to managing live models, frequent retraining, and modular AI components.
The OG “-ilities” remain the backbone of resilient system design—but the AI era requires that we layer on new expectations for transparency, responsibility, and ethical alignment.
What’s your take? Are there “-ilities” for AI you think should be front and center? Let’s keep evolving this cornerstone of software quality together.