Blog Series: QA Reimagined – Navigating Agile, AI and Automation-Part 5

Part 5: AI in QA – Beyond the Buzzwords, Into the Real Work

Previously in this series:

In Part 4, we explored how practices like feature flags and trunk-based development shift the way we think about testing.

The message was clear: QA isn’t a checkpoint anymore — it’s a constant, contextual presence.

Now for the topic we have all been waiting for AI, Agents and LLMs. This post picks up from the past few posts, intentionally — but adds a new dimension:

How can AI actually support QA today — in ways that are helpful, not hype? It’s more to show that AI in QA does not just come out of the blue, it comes as an evolution to what we already know, cultivated and built as our craft and area the we love so much-QA Engineering.

AI in Testing: We’ve Moved Past the “Wow” Phase

Let’s be honest: the past year has been filled with “AI + testing” announcements — auto-generating test cases, writing scripts from prompts, autonomous testing agents, etc.

And while it’s exciting, I’ve found most teams are still asking:

“But how do we use this in practice — with our codebase, our tests, and our people?”

In this post, I’ll focus on the real use cases I’ve seen working. These aren’t sci-fi. They’re practical, targeted, and most importantly — they fit into how teams already work.

Where AI Actually Helps (Today:2025 )

1. Test Case Generation from Requirements

With LLMs, we can now generate:

-Gherkin-style scenarios from user stories

-Acceptance test outlines from Jira tickets

-Manual test case drafts from Confluence pages

This helps teams get unblocked faster, especially when QA joins mid-sprint or when specs are vague.

Tip: Always validate — but don’t start from scratch.

2. Root Cause Analysis (RCA) Acceleration

LLMs can:

Read failed test logs Cross-reference recent code changes Suggest likely failure areas

I’ve seen this help triage complex integration failures within minutes — saving hours of Teams/Slack messages and blame-chasing.

Think of this as a co-pilot for defect analysis, not a replacement for engineering intuition.

3. Smart Test Selection and Prioritization

Tools or internal LLM-based RAG setups help answer:

Which tests should run based on this commit? Can we skip tests unrelated to the change?

This matters more than ever in monorepos, microservice-heavy environments, or cross-platform suites.

The goal isn’t fewer tests — it’s smarter pipelines that give faster feedback.

4. Accessibility and UI Heuristics

LLMs and agents can now:

Scan UI designs or web apps for accessibility violations Flag common anti-patterns (e.g., missing labels, poor contrast) Offer contextual suggestions inline

Early stages, yes — but a great tool in exploratory testing, especially for WCAG compliance.

5. Exploratory Testing Assistance

During manual testing sessions, AI copilots can:

Record user actions and annotate test steps Suggest variations or edge cases to try next Capture screenshots + logs automatically

I’ve tested this with mixed results — but it’s promising for teams trying to elevate the quality of exploratory work.

The value isn’t just in AI automation — it’s in lowering friction.

What AI Still Can’t Do (Well) *based on my viewpoint*

Let’s be clear — this isn’t magic. AI in QA still struggles with:

-Domain-specific logic (unless trained with your own data)

-Flaky test identification (unless signal-to-noise ratios are high)

-Deep exploratory intuition (that human curiosity edge we all have)

-Systems that require understanding non-textual complexity (multi-modal flows, hardware interactions)

Use AI to assist your judgment — not to replace it.

My Guidance for QA Leaders and Teams* very important golden rules

If you’re trying to introduce AI into your QA practice:

a) Start with pain points. Where are you losing time — writing boilerplate, triaging defects, checking logs?

b) Use your existing data. Connect LLMs to your Jira, Test Management tools (Xray/Test Rail etc), test logs, or commit metadata. Don’t expect good answers from general prompts.

c) Focus on augmentation. AI is your accelerator, not your decision-maker.

d) Build trust slowly. Validate its outputs. Refine the prompt. Re-evaluate what success looks like.

e) Train the team. Not just on tools — but on what to expect, how to review AI output, and when to trust vs override.

Closing Thoughts: Practical, Not Just Promising

In this post I purposefully kept it at more of a strategic level related to AI, there is a lot more under the surface from different types of LLMs, MCPs and advanced agents, but that’s for a whole new blog series… watch this space

We’re at a moment where AI can actually support real QA work. But it requires alignment with reality: Your context , Your tests , Your pain points, Your people

The good news? The tools are ready.

The challenge? The thinking has to catch up.

The best AI solution in testing is the one that fits your team’s needs, not someone else’s roadmap.

Up Next:

I conclude the blog series with some reflections, predictions and forecasts that may be the differentiator in what the future brings.