Will future QAs need more empathy than code?
During a recent interview with a well-known generative AI company (NDA signed, so I can’t name names), I encountered something unusual. After a strong first round with the CTO, my second interview wasn’t with Product or Engineering — it was with the head of Customer Support.
Most of the questions weren’t about automation, frameworks, or test strategies. They were about collaborating with support teams: filing bugs from user complaints, interpreting ambiguous cases, and turning customer frustrations into actionable reports. I remember scrambling to connect my limited experience with support to the questions they asked, while internally wondering: Why is QA sitting so close to support in this company?
Spoiler: I didn’t get the role.
At first, I dismissed it as a “weird” interview. But soon, I noticed a pattern. Many AI companies — especially those building generative systems like chatbots or models similar to GPT — are advertising QA roles where support collaboration comes first, automation second (if at all). For instance, here’s a requirement list from Notion’s AI QA role:
- Experience in Customer Experience or QA, analysing support responses at scale
- Strong written communication, turning insights into actions
- Collaboration across CX, Ops, and vendor teams
Notice something? Automation isn’t even mentioned.
This closely mirrors my interview experience. The emphasis is not on automation at all, but on working with Customer Support and having the ability to “turn insights into clear actions.” The way I see it, among other things, that means taking user complaints and helping to turn them into actionable bug reports.
What’s going on here? I don’t have hands-on experience in testing AI chatbots — so, ironically, I asked one, to help me understand this better. The insights I uncovered made everything click, and honestly, it was eye-opening.
So…
Why QA Looks Different in Human-Facing AI
Quality is perception, not just correctness.
An answer can be factually right but come across as rude. A recommendation can be relevant but still feel useless. These nuances often surface through customer complaints — not crash logs.
Edge cases emerge from real users.
No amount of pre-release testing will uncover every strange prompt, unusual accent, or unexpected input. Support sees them first, and QA must work with them to decide: bug, limitation, or acceptable quirk?
Faster triage depends on support.
Many AI failures don’t break systems — they confuse users. Support logs these, QA interprets and prioritises, and engineering decides what to fix. QA acts as the translator in this cycle.
What this means?
QA roles are evolving to prioritise curiosity, empathy, and user centered thinking.
In testing chatbots or generative AI, traditional tools — test cases, reproduction steps, and even automation scripts — often take a back seat. Our most valuable skill remains our curious, experimental nature, and for testers entering AI, mastering empathy and user-centric thinking may soon matter more than perfecting automation scripts.