The topic of AI came up a fair bit at the recent LEAP 2.0 API governance conference. Perhaps unsurprising, given how many ways organizations are trying to embrace its uses, as they race to benefit more from the AI revolution than their competitors. But what impact is AI having on the working models of platform engineering teams?
Leading tech journalist Jennifer Riggins shared some thoughts on this at the Tyk conference, so we’ve shared her top takeaways below, including:
- Why organizations need to rethink the problems they’re trying to solve with AI
- Why API governance best practices need to apply to AI governance
- How better communication can lead to more effective use of AI
What are we trying to solve with AI?
A lot of organizations are spending a lot of money on AI tools but many of them aren’t clear on the problems they’re trying to solve with AI. For example, if half of AI-written code is buggy, and writing code is what makes developers happy and productive (a topic that this Microsoft study dives into), why are organizations trying to use AI to automate coding?
Instead, it might be better to focus AI on solving problems that developers have self-identified, with documentation and technical debt being chief among these. AI is well-placed to deliver on both fronts. It’s good at explaining complex topics and understanding complex things like code bases where nobody knows who owns what. It’s very helpful with API discovery. And it’s great at creating documentation that is searchable in a natural language.
How do we build best practices into AI usage?
There’s also the matter of specification-driven development, or behavior-driven development, to consider. This is where you use plain language to describe new functions, develop code based on those specifications, then test the code against the spec. This is an API best practice that needs to happen in the generative AI space. Yet fewer than half of organizations have an AI policy; nothing to tell staff not to put customer information, human resource information, proprietary code and more into AI tools.
AI can certainly support spec-driven development, though with humans generating the code. We just have to ensure that API governance best practices are also applied to AI governance. Doing so can improve consistency and embrace earlier validation that what an organization is building is what it should be building.
Do we have a technology problem here or a people problem?
What we really have is a people problem – a communication problem. There’s a chasm between the business and the technology. Businesses need better insight into what’s going on, with better communication with stakeholders in natural language. That’s the point at which we can start applying generative AI to tasks that it does well and that are actually solving developer problems.
Interested in finding out more? Then why not read Tyk CEO Martin Buhr’s thoughts on AI governance in action?
