# What the World Quality Report 2025 and DORA 2025 Say, and What a QA Lead Does on Monday

Dokimore Admin Team · October 28, 2026, updated 2026-11-10 · 6 min read

Format: Report read · Written for: QA leads and test managers, Engineering managers · Where it applies: every market

> Eighty-nine percent of organisations are piloting or deploying generative AI in quality engineering, and fifteen percent have scaled it. The gap between those two numbers is the year’s work.

## In short

- The World Quality Report 2025 (17th edition, November 2025): 89% of organisations are piloting or deploying generative AI in quality engineering, 37% in production and 52% in pilots, but only 15% have scaled it enterprise-wide, 43% are still experimenting, and the share with no adoption rose to 11% from 4% the year before. The barriers are data privacy (67%), integration complexity (64%), hallucination and reliability (60%) and skills (50%).
- DORA 2025 (September 2025, about 5,000 respondents): AI adoption reached 90%, 65% rely on it heavily, and adoption is now linked to higher delivery throughput, a reversal of 2024. The trust is thin, 24% trust AI a great deal or a lot against 30% a little or not at all, and the report’s framing is that AI amplifies what an organisation already is.
- Read together, the two reports say the tooling is in and the system around it is not: a QA lead’s Monday is spent on the validation strategy, the test data and the evidence trail that let a pilot become a practice, not on another pilot.

## The two reports, and what they are

The World Quality Report is the annual survey of quality engineering from OpenText, Capgemini and Sogeti; the 2025 edition is the seventeenth, published on 13 November 2025 from more than 2,000 senior executives across 22 countries and ten sectors. DORA’s report is Google Cloud’s annual research on software delivery, which in 2025 was titled State of AI-assisted Software Development and drew on a survey of about 5,000 technology professionals. One asks executives about quality programmes; the other asks practitioners how they work. That difference explains most of where they diverge.

*The two reports side by side, in the figures this article quotes*

| Question | World Quality Report 2025 | DORA 2025 |
| --- | --- | --- |
| Published by | OpenText, Capgemini and Sogeti | Google Cloud’s DORA research |
| Who answered | More than 2,000 senior executives in 22 countries and ten sectors | About 5,000 technology professionals |
| Generative AI use | 89% piloting or deploying; 15% enterprise-wide | 90% use AI for development; 65% rely on it heavily |
| The gap it names | Scaling: data privacy 67%, integration 64%, reliability 60%, skills 50% | Trust: 24% trust AI a great deal or a lot, 30% a little or not at all |
| Productivity | An average reported gain of 19%; a third see minimal gains | More than 80% report a gain; adoption now linked to higher throughput |

## What the World Quality Report found

Adoption is broad and shallow. 89% of responding organisations are piloting or deploying generative-AI-augmented workflows, with 37% in production and 52% in pilot. Only 15% report enterprise-wide implementation; 43% remain in the experimental phase and 30% operate within limited use cases. The share of non-adopters rose to 11% from 4% in 2024, which the report notes is still well below 2023’s 31%: some organisations tried it and stepped back.

The returns are uneven. Organisations report an average productivity gain of 19%, and one third have seen minimal gains. The barriers to scaling are data privacy risk (67%), integration complexity (64%), hallucination and reliability concerns (60%) and a lack of AI and machine-learning expertise (50%, unchanged from 2024). Shift-left remains the dominant approach in quality engineering, and shift-right, testing and observing in production, is gaining ground.

## What DORA found

Adoption is near-universal among practitioners: 90% use AI for software development, a 14% increase on the year before in the report’s words, and 65% rely on it heavily. More than 80% say it improved their productivity and 59% say it improved code quality. The headline change from 2024 is that AI adoption is now linked to higher software delivery throughput, which reverses last year’s finding; the post that announced the report names the remaining challenge as ensuring that software works as intended before it is delivered to users.

Trust lags use. 24% of respondents trust AI a great deal or a lot; 30% trust it a little or not at all. The report’s own framing for the whole picture is that AI is a mirror and a multiplier: in a cohesive organisation it boosts efficiency, in a fragmented one it exposes the weaknesses that were already there. It offers a companion, the DORA AI Capabilities Model, with seven capabilities that separate the teams that benefit from the teams that merely adopt.

## Where they agree, and the one place they do not

They agree that the tool is in and the practice is not. The executives’ 15% scaled against 89% piloting or deploying, and the practitioners’ 90% use against 24% high trust, describe the same organisation from two floors of the building. They agree on the barrier, too: the executives call it data privacy and integration, the practitioners call it trust, and both are the absence of a system around the tool, validation, data handling, evidence, that would let anyone rely on its output.

They do not agree on productivity. More than 80% of DORA’s practitioners say AI improved their productivity; the World Quality Report’s executives report an average gain of 19%, and a third of them minimal gains. Both can be true: a developer saves an hour and the organisation ships no faster because the hour moves to review, rework and the test that nobody wrote. DORA’s own throughput finding, higher throughput with the works-as-intended problem still open, is the same story told the other way round.

## What Dokimore has to do with any of this

Three of the five Monday items are evidence, test data and a monitor, which is the shape of the product: every recorded run keeps the element, the network calls and a screenshot per step; the test-data generator produces seeded, locale-real synthetic data with checksum-valid but fictitious identifiers; API collections run on a schedule as monitors with alerts. The platform’s own analysis runs on local models, and nothing goes to an outside AI provider unless you connect your own key in the LLM testing module or your team buys and switches on Work with AI, which is one answer to the privacy barrier. None of that writes the validation strategy; that is the page a QA lead writes on Monday morning.

## What a QA lead does on Monday

1. Write the validation strategy for one AI-assisted workflow, on one page: what the tool produces, who checks it, against what, and what evidence the check leaves. The 2024 World Quality Report named the lack of a validation strategy as the top barrier, and the 2025 barriers are its descendants.
2. Replace production data in the test environment with synthetic data. Data privacy is the top scaling barrier at 67%, and a pilot that touches real customer records does not survive the review that turns it into a practice.
3. Make the evidence per run the artefact, not the pass rate. A run that keeps what each step did is what lets an executive trust a number a practitioner produced with a tool.
4. Pick one shift-right measure: a scheduled monitor on the three journeys that earn money. The report says shift-right is gaining traction; the reason is that it finds the failures the tool-assisted suite was never asked about.
5. Count the AI-assisted changes that reached production last month and the incidents in the same month, and try to connect them. If the connection cannot be made, the gap is the evidence trail, and it is the cheapest thing on this list to fix.

## Questions people ask

### What does the World Quality Report 2025 say about AI in testing?

That adoption is broad and shallow: 89% of organisations are piloting or deploying generative-AI-augmented workflows, 37% in production and 52% in pilot, but only 15% have scaled it enterprise-wide and 43% are still experimenting. The average reported productivity gain is 19%, a third see minimal gains, and the top barriers to scaling are data privacy risk (67%), integration complexity (64%), hallucination and reliability (60%) and skills (50%).

### What did DORA 2025 find about AI and software delivery?

From about 5,000 technology professionals: 90% use AI for development, 65% rely on it heavily, more than 80% report a productivity gain, and AI adoption is now linked to higher delivery throughput, reversing the 2024 finding. Trust is thin, with 24% trusting AI a great deal or a lot and 30% a little or not at all, and the report frames AI as a mirror and a multiplier of what an organisation already is.

### Who publishes the World Quality Report?

OpenText, Capgemini and Sogeti, annually. The 2025 edition is the seventeenth, published on 13 November 2025, from a survey of more than 2,000 senior executives across 22 countries and ten sectors.

### Is shift-left still the main approach?

The World Quality Report 2025 says shift-left remains the dominant approach in quality engineering while shift-right, validating and observing in production, is gaining traction. The practical form of shift-right for a small team is a scheduled monitor on the journeys that matter most, which catches the failures a pre-release suite cannot see.

## How we measured this

- Every figure is quoted from the publisher’s own press release or announcement post, read on the date beside the reference; we have not seen the underlying survey data and we have not reproduced either study.
- The Monday list is our reading of the two reports, not a recommendation either publisher makes.
- Corrected 2026-11-10: the article said nothing goes to an outside AI provider unless you connect your own key in the LLM testing module; since 1 November 2026 the Work with AI add-on, for a team that buys it and switches it on, is a second way.

## Sources

1. [Capgemini press release, "World Quality Report 2025: AI adoption surges in Quality Engineering, but enterprise-level scaling remains elusive" (13 November 2025): 17th edition, 2,000+ executives in 22 countries, 89% / 37% / 52% / 15% / 43% / 11%, 19% productivity, barriers 67% / 64% / 60% / 50%](https://www.capgemini.com/news/press-releases/world-quality-report-2025-ai-adoption-surges-in-quality-engineering-but-enterprise-level-scaling-remains-elusive/) (read 2026-10-28)
2. [Google, "How are developers using AI? Inside our 2025 DORA report" (23 September 2025): about 5,000 respondents, 90% adoption, 65% heavy reliance, throughput reversal, 24% / 30% trust](https://blog.google/technology/developers/dora-report-2025/) (read 2026-10-28)
3. [DORA, "State of AI-assisted Software Development" report page: AI as an amplifier; the DORA AI Capabilities Model](https://dora.dev/research/2025/dora-report/) (read 2026-10-28)

Source: https://dokimore.com/blog/world-quality-report-2025-and-dora-2025-the-qa-leads-read
