Harvendra Singh, IT Delivery Manager - Cloud Engineering & Architecture, driving innovation through cloud and AI-led digital transformation.
getty
Cloud platforms are starting to think for themselves.
Well, not quite, but AI is being injected into cloud platforms at every layer. As applications become more dynamic, adaptive and data-driven, there is an incredible opportunity to do things we’ve never dreamed possible. But it also begs the question: How do you test and secure systems that learn and adapt over time?
The old guard of quality assurance (QA) was built around stable, predictable systems. Inputs were well-known, outputs were deterministic, and if you threw a monkey at your application, it wouldn’t break. That’s not the case anymore. Modern systems are event-driven, respond to real-time inputs and act across distributed services.
It’s time we started treating testing like the complex problem it is.
Modern software systems are complicated. While most organizations acknowledge that, they often underestimate the complexity involved. We’re talking about millions of events per minute, integrated across dozens of services and changing every day as new data flows into the system.
When you introduce machine learning models, it only gets worse. The same input can trigger different results based on context. Couple that with the distributed nature of cloud systems, and validation becomes a moving target.
I’ve seen it happen during large-scale cloud transformations. Teams take their monolithic application—something they understand well—and start to modernize it. Everything looks good at first, but problems soon surface.
“How do we know if everything will work in production?”
There isn't a single answer to that question because your system doesn’t behave the same way twice.
If your quality control process is built around checking things off of a checklist, it’s time to evolve.
Modern applications are event-driven, meaning the same input can trigger different outcomes depending on context. Add machine learning into the mix, and you get outputs that might be right but for the wrong reason.
There’s no longer a clear-cut “wrong” answer when validating AI-powered applications. Testing must be continuous, contextual and automated.
So, how do you test systems that are continuously learning and evolving?
Instead of manually scripting tests, we can automate test case generation with AI. Some teams are experimenting with prompt-driven testing, using models to generate tests automatically.
For example, you can use a large language model to generate tests based on real-world inputs. When combined with structured context in the form of a knowledge graph, you end up with far more meaningful test generation.
There’s already research demonstrating this approach. Prompt-driven test generation using large language models and knowledge graphs shows how testers can improve coverage and accuracy in data-intensive systems.
The takeaway is simple: Testing must evolve alongside system complexity.
Chances are you’re not going to get perfect test coverage, and that’s okay.
If your system behaves differently based on context, your testing strategy needs to adapt. That means testing in real time.
This typically includes:
• Monitoring: Capture signals to observe system behavior.
• Test generation: Continuously create new tests based on patterns.
• Alerting: Identify anomalies as they happen.
• Validation: Confirm outcomes are correct.
On a recent engagement, we built a simple pipeline that ingested system signals in real time and connected it to a lightweight model that flagged anomalies and triggered automated responses. It wasn’t perfect, but it caught issues minutes after they occurred instead of hours or days.
Sometimes being just a little bit smarter is all it takes.
As AI becomes more embedded in cloud systems, testing and security start to converge.
Why? Because once systems begin making decisions automatically, they will behave in ways you don’t expect.
Intelligent testing helps identify:
• Errant model decisions
• Model drift
• Distributed system failures
• Edge cases missed by traditional testing
By embedding testing into the system itself, you move from reactive validation to proactive assurance. Instead of hoping everything works, you gain confidence that your system behaves as intended.
Transformational technology rarely happens all at once. Cloud computing didn’t. Containers didn’t. And neither will intelligent testing.
The most effective approach is incremental. Start with one critical path, and introduce intelligent testing where it matters most.
Rather than replacing your entire QA process, look for opportunities to improve step by step. Small changes, applied consistently, create meaningful impact.
Applications will only continue to grow more complex as AI becomes more integrated into cloud platforms.
The question isn’t whether systems can scale, it’s whether they can be trusted.
How will you ensure your systems are secure? How will you validate behavior under constantly changing conditions?
Don’t wait for perfect automation. Start small, and introduce intelligence where you can.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。