Industry outlook

The direction of travel is broader AI participation in delivery. The execution problem is making that participation reliable, integrated and economically useful.

Analyst perspectives: strategy and organizational change

Gartner’s public forecasts anticipate much wider assistant use and changes in team structure. Its public testing research abstracts emphasize platform choice and continuous quality. These indicate where the analyst sees demand; they do not establish adoption outcomes or reproduce the licensed research. [G01][G02][G03][G04]

McKinsey’s software-development research places AI within the wider product lifecycle and operating model. Its 2026 bank example illustrates an ambitious implementation direction, but the public account does not supply enough independent detail to make the claimed gains a planning baseline. [M01][M02]

Implication for an executive: define the future capability and how it changes delivery. A license rollout alone leaves the integration, verification and ownership work unresolved.

Experimentation is ahead of enterprise scale

The World Quality Report reports 89% piloting or deploying GenAI-enabled QE. In a separate maturity breakdown, 15% report enterprise-wide deployment. These are answers to different questions and should not be joined into a single conversion funnel. [W01]

What constrains adoptionShare of respondents reporting each barrier · World Quality Report 2025–26
Data privacy
67%
Integration complexity
64%
Hallucination / reliability
60%
AI / ML skill gaps
50%

Overlapping responses, not shares of a whole. Survey: >2,000 executives; question-specific n unavailable in the public release. Source: Capgemini, Sogeti and OpenText, November 2025. [W01]

The barriers suggest a practical investment agenda: approved data access, delivery-system integration, evidence of reliability and staff capability. Buying a more capable model does not itself complete that agenda.

Quality work extends beyond test execution

DORA describes AI as interacting with the surrounding engineering system. Its 2026 qualitative study also draws attention to the effort required to understand, verify and correct generated work. [D01][D02]

For QE, our synthesis is a shift in emphasis:

Existing capability Added AI-era responsibility
Requirements and acceptance tests Explicit task contracts and machine-checkable expectations
Test automation Evaluate generated artifacts and protect test validity
Defect triage Diagnose failures with traceable context and calibrated uncertainty
Release assurance Assess model, prompt, retrieval and tool changes together
Production quality Curate failures and drift into repeatable evaluations

This is an expansion of engineering responsibility. Staffing decisions should follow measured changes to work, service levels and risk, with a deliberate plan to develop less-experienced engineers.

Read next: Evidence and economics explains why reported benefits differ.