Deepa Subramanian

Code-based testing has gone autonomous. Non-code-based testing? Not so much. Our CEO, Dan Faulkner, breaks down why that needs to change, and what it means for the future of software quality. Check it out 👇

2

I’m seeing a lot of posts about AI in Quality Engineering right now - what it means, how it changes things, where it fits. Here’s my take. We’ve been in an enablement mindset long before AI showed up. When automation accelerated, we adapted. When shift left became the focus, we moved earlier into design, refinement, and architecture conversations. Each time, the role evolved and we moved with it. Automation changed how we test. Shift left changed how and when we engage. AI is now accelerating how fast we learn and generate feedback. But the key factor is quality engineering has already been shifting from execution to enablement. Designing feedback loops isn’t just a quality engineering task. Sometimes it’s a developer introducing contract testing. Sometimes it’s observability built into services. Sometimes it’s better CI pipelines. Our role is to help teams identify risk early and place feedback loops where it creates the most leverage. AI doesn’t create that mindset - it amplifies it. Quality today is not just about running tests and more about driving conversations: Where is the risk? Where should feedback live? How fast can we learn? The tools evolve. The responsibility to think systemically doesn’t.

24

3 Comments

Being on Joe Colantonio's TestGuild was very much a career highlight for me. Getting to join a second time, pheww pinch me. My takeaway I'm hoping you get from this: what if understanding the codebase was no longer a blocker to great testing? Historically, testers were forced to work around the code, clicking through UIs, guessing selectors, relying on outdated docs, or waiting on developers for answers. The conversation flips that model on its head. With tools like Cursor, testers can now interrogate the codebase directly by asking questions in plain language. We walked through real-world examples showing how testers can: Explore APIs, data models, and relationships without digging through files manually Understand what changed in a release before writing a single test Identify risk earlier and design smarter test strategies https://lnkd.in/g3yiPWXH Generate Playwright tests faster, with more context and confidence Shift from writing everything by hand to reviewing and refining high-quality first drafts This isn’t about turning testers into developers or handing control to “AI agents.” It’s about using AI as an information accelerator, a way to finally see how systems actually work, reduce low-signal work, and focus human judgment where it matters most. If you’re onboarding to a legacy codebase, struggling to understand impact from new changes, or tired of guessing what to test, this episode is for you. Highly recommend checking it out, and if this resonates, I’ll also be running a 90-minute hands-on workshop at Automation Guild 2026 where we go step by step through these techniques in practice.

46

3 Comments

💻 “Not Technical Enough?” – Let’s Talk About Bias in Hiring 💻 Ever seen a female engineering candidate passed over for “not being technical enough” despite an impeccable track record and a portfolio full of complex, high-impact technical achievements? I have—and it’s a pattern that says more about the process than the person. This isn’t about ability. It’s about bias: • Different standards of proof – Men are often assumed competent until proven otherwise; women are asked to re-prove competence at every step. • Invisible criteria – Vague feedback like “not technical enough” hides subjective judgments and makes growth impossible. • Role framing – Women are more likely to be steered toward “soft skills” or leadership, which can be misread as lack of technical depth. 💡 Call to action Hiring teams can change this by: • Using structured, skills-based interviews and rubrics. • Defining “technical” with clear, measurable expectations. • Valuing diverse technical paths—architecture, infrastructure, frontend, data, DevOps, and beyond. Technical excellence has no gender. It’s time our hiring practices reflected that. #WomenInTech #SoftwareEngineering #InclusiveHiring #BiasInTech #Leadership

2

The AI productivity conversation has me wondering about a potential 'Big Uh-Oh' hiding in plain sight. Not taking sides here... just surfacing a tension I've seen in debates across QA, dev, and exec teams (customers as well as groups here on LinkedIn). Curious what y'all think (I know, a New Englander using "y'all"... cut me some slack)? The Root Tension: Production just moved to machine speed. Verification is still running at human speed. That's not acceleration... that's a high-velocity bottleneck. Your AI is generating, say, maybe 10x the output. But every unit still needs a human to determine if it's trustworthy. The review queue doesn't shrink because the tool is faster. It explodes. AI is the accelerator... a reinforcing loop driving volume up fast. Human verification is the balancing loop... the system's natural stabilizer. When those two forces fall out of balance, the system doesn't just slow down. It fails. The "Teenage Son" Analogy: My colleague's sage wisdom about his son comes to mind: "I trust him & I raised him right. But he's a teenage boy... they're inherently untrustworthy. So I love him, I trust him... and I verify the big things. Not because he's bad. Because he's unproven." To me it's a near perfect analogy. AI is exactly the same... well ... for now. Both will mature over time, but unproven systems require discipline, not just hope. QA Discipline: In my QA life, I run regression tests even when I expect zero bugs. You change the system, you validate the system. That's not distrust. That's discipline. The Balancing Loop Challenge: Corporate response right now seems to be cutting their "verification people" indiscriminately. Removing the "Human Hardware" exactly when production load is spiking. That's not tuning the balancing loop. That's disabling it. and before you say "just use a second AI to check the first one"... that's not balancing. That's two accelerators agreeing with each other. One machine validating another's "plausible enough sounding" output isn't verification. It's an echo chamber. AI can tell you if the code is well-formed. It can't tell you if it's right for the business. The machine knows the prompt; the human knows the consequences of the prompt being wrong. My Questions: Is the goal of sustainable AI productivity with certainty mean fewer humans or more? Lowering unit verification cost? Better craft? Smarter tooling? Architectural constraints that reduce what needs reviewing? Is anyone actually investing in the balancing loop (and How?)... or just riding the accelerator hoping nothing breaks? I am honestly seeking clarity & discussion here. I'm not against the accelerator. I just want to be cautious about a car with no brakes (every LinkedIn post needs a cliché, right?).

6