Humans in AI: the unseen backbone of quality engineering

In an era where AI promises autonomy, the World Quality Report 2025–26 (WQR) reminds us of a critical truth: humans remain the cornerstone of trust, ethics, and quality assurance in AI-driven ecosystems.

Generative AI has crossed many boundaries, and today’s applications are undeniably impressive. From investing real money in stocks, managing peak loads during events like Giving Tuesday or Black Friday, to powering robotics at the edge, the possibilities seem infinite.

It’s an exhilarating time for technology- thrilling, adventurous, and sometimes reckless. The rapid advance of AI can feel like humanity is being lifted to the next level. Yet, this very excitement is why we must pause and reflect. The sensation of stepping into the unknown with a tool as powerful as AI is intoxicating, but it demands caution and mindfulness.

Experimentation is essential- think of Dolly the sheep, whose cloning pushed biotechnology forward. But ethical concerns and the potential impacts were carefully considered, and the technology did not disrupt everyday life at scale. In contrast, GenAI is scaling rapidly and permeating daily life, often without sufficient checks and balances. While it brings positive outcomes, it also enables new forms of fraud and unethical practices.

In enterprise applications, AI is ushering in a new era for IT—transforming both the Software Development Life Cycle and business operations. However, the nature of AI-based applications means it’s dangerous to leave them unsupervised. Consider these real-world examples:

• AI writing emails for a start-up owner without consent.
• Dropping a database despite explicit instructions not to deploy certain code.
• An airline forced to honour a refund policy mistakenly described by its GenAI-powered chatbot.
• A financial recovery chatbot responding on the lines of “Ok, please do that, I will wait” when a customer mentioned selling a kidney to pay back a loan.

These cases highlight that AI cannot always be trusted to do the right thing or follow instructions precisely. They raise important questions about how we embed ethics and accountability into the applications we create.

While the excitement about AI’s potential is justified, it’s the focus on underlying issues that will make AI viable and scalable in real-world scenarios. Quality Engineering (QE), hence, has a pivotal role to play.

As user stories shift from simple pass/fail criteria to fuzzy boundaries, human experts in the loop become central to the operation of AI-infused applications. Scaling these solutions requires the entire ecosystem-leadership, test data, skillsets, domain knowledge, culture-to mature in tandem.

Humans form the backbone of trust, ethics, accountability, and resilience in Quality Assurance. This isn’t just our view—the World Quality Report reflects this sentiment from over 2,000 survey participants. Here are some key findings from the WQR that reinforce the human role in AI-driven QE:

Top WQR findings: humans in AI & quality engineering

1. Human accountability for AI outputs is non-negotiable.
   QE teams remain fully responsible for every test case, decision, and outcome—even when AI generates artefacts. Human experts must critically evaluate and challenge AI outputs rather than accept them at face value.
   Source: Quality Engineering in AI (Chapter 1), pp. 20–23; “The expert in the loop” & challenges incl. accuracy/hallucination at 60%; Fig. 5 & narrative.
2. Upskilling is the human lever: skill gaps stall adoption.
   Only 53% of testers have upskilled in AI/ML; 46% of organisations have created AI-specific QE roles. WQR calls for structured training with validation to prove the ability to challenge AI outputs—not just operate tools.
   Source: Quality Engineering in AI (Chapter 1), Fig. 4 & recommendations, pp. 22 & 27–29.
3. Humans set the right scoreboard: metrics must reflect impact.
   Only 25% tie QE metrics to business outcomes; most still track activity (defect counts, coverage). WQR urges human-chosen impact metrics (release predictability, defect containment, adoption, revenue/risk reduction).
   Source: QE in Agile (Chapter 4), Fig. 31 & narrative, pp. 60–62; Chapter 1 recommendations, pp. 27–29 (“Focus metrics on transformation impact”).
4. Domain knowledge is the multiplier for AI value.
   WQR shows domain expertise rising in importance (e.g., 56% in Agile skills ranking). AI amplifies existing expertise rather than replacing it; without business/regulatory context, AI artefacts are brittle.
   Source: QE in Agile (Chapter 4), Fig. 25, pp. 57–59; Enterprise QE (Chapter 5), skill prioritisation Fig. 36, p. 69–71.
5. Data quality & stewardship: a human craft behind AI.
   Despite 95% using GenAI for test data, only 10% achieve full lifecycle integration. Persistent issues—poor data quality (51%), accuracy gaps (48%), dataset scaling (49%)—reflect the need for human data stewards and governance.
   Source: Data Quality / TDM (Chapter 3), Fig. 17–24 & narrative, pp. 23–27 & 47–53.
6. Resilience requires human strategy, not just observability.
   Shift-right adoption is tool-first; chaos testing and resilience engineering are underused. WQR urges human-led resilience strategies that connect production telemetry to design/testing via closed loops.
   Source: Shifting Quality Right (Chapter 6), Fig. 39–44 & narrative/recommendations, pp. 39–42, 79–83.

For more information and a detailed read out – please download the report from here

In summary:
AI may be the engine, but humans are the drivers—ensuring trust, ethics, and quality remain at the heart of every innovation. As the World Quality Report 2025–26 makes clear, the unseen backbone of Quality Engineering is, and will remain, human.