I asked a senior QE leader three questions about AI, automation, performance testing, and what recruiters should actually be screening for.
Here’s what he had to say.
1. AI Agents in the QA Workflow
Q: What AI/automation initiatives are you focused on, and what problems do they solve?
A:
We’re building a set of AI agents into our QA workflow that handle specific, well-scoped tasks – identifying what a new requirement needs tested, keeping the existing test library current as requirements change, telling flaky failures apart from real defects and assembling evidence for release-readiness decisions instead of relying on a meeting and someone’s memory. The common thread is consistency: judgment calls that used to live in one engineer’s head now get made the same way every time, with a trail behind them. Alongside the new capability, we’re just as focused on the guardrails – agents propose, humans decide and there are clear limits on what an agent is allowed to touch directly in our systems of record.
2. AI and Performance Testing
Q: How is AI changing performance testing?
A:
The metrics we care about haven’t changed: response time, throughput, scalability, resource use. What’s changing is targeting and triage. AI now helps flag which changes are likely to need performance coverage, so that decision doesn’t depend on someone remembering to ask. It also does a first pass on failures – sorting environmental noise from a genuine regression – before an engineer looks. That’s where it’s helped most: narrowing what a human needs to focus on, not replacing their judgment. The higher stakes, full-scale load testing stays deliberately human run. We’re intentionally not putting a hard number on time savings yet because we’d rather get the model right than publish a metric we don’t fully trust.
3. What Recruiters Should Be Screening For
Q: What skills should recruiters be screening for that they’re missing?
A:
Less AI tool fluency than people assume and most candidates have that. More: judgment about where automation should and shouldn’t have final say, real depth working across multiple integrated systems rather than one clean API, and the one I think gets missed most – skepticism toward AI output rather than enthusiasm for it. I’d rather hire someone who can describe a time they caught an AI system being confidently wrong than someone with a longer list of AI projects.