Systems | Development | Analytics | API | Testing

According to SmartBear’s Closing the AI Software Quality Gap study, 93% of teams are already using AI to generate code. The same study found that 60% expect AI to produce nearly half of all code within the next year. This shift in development velocity is already impacting software testing and quality. Most teams say application quality is suffering, and 60% have experienced quality issues in the past year because development is moving faster than testing can keep up.

Customers today face significant challenges as their Kubernetes environments scale. The proliferation of microservices, external integrations, and new AI workloads increases traffic volume and connectivity complexity, creating material risks to performance and availability. The core issue is a lack of end-to-end governance: as diverse workloads expand, unmanaged interactions make it difficult to apply consistent security and enforce global consumption policies.

RESTful APIs and microservices solve different problems — REST is a style of API design, microservices is a pattern for structuring an application — but they work together so often that they're frequently confused. Most production microservices architectures use REST as the default communication mechanism between services, while plenty of monolithic applications also expose RESTful APIs to external clients.

Your AI infrastructure is producing financial signals your CFO has never seen. Token consumption is a direct cost line item. Cache hit rate is a margin improvement. Model routing decisions are cost arbitrage events. These things are happening right now, in the gateway layer, with no route to the CFO, which means no route to the boardroom. As the AI connectivity platform owner, you're the person who can build that route.

There's a reason we don't find our keys by scanning every room like a security camera. We replay the tape. We remember the groceries, the front door, the distraction. We reconstruct the *why* to find the *where*. Our brains are commit logs, not snapshots. Most agentic AI systems today work more like the camera — a static frame of the world at a given moment. They store state. They retrieve context. They produce an answer.

We’ve been building an AI agent that can take a production bug, find the root cause in captured traffic, write a fix, and validate it before a human reviews it. We call it Agent Factory. Last week we ran it on ourselves, against a real bug in our own production service. The first thing we did was get the workflow wrong.

This is Part 2 of the AI Software Factory series. In Part 1, we established that the Agile methodology is buckling under the weight of “elastic code.” When AI agents can generate functionality in seconds, two-week sprints and manual task management become organizational bottlenecks. We introduced the concept of the AI Software Factory: a shift from managing human tasks to managing business intent through a “Funnel of Increasing Trust.” But a factory requires infrastructure.

In an API-driven setup, a gateway often sits between clients and backend services: it can validate input, aggregate upstream responses, and give you one place to observe traffic. Koa is a strong fit for that role. Its core stays small, async/await is first-class, and middleware composes in a predictable stack. In this article, you will build a compact API gateway with Koa that: You will also wire up AppSignal for the Node.js stack.