News

The AI-Native Developer: Inside Google Cloud's 2025 DORA Report

• By John K. Waters
• 09/23/2025

In 2025, software development no longer happens without AI. What began as an optional coding assistant has become a near-universal presence in the developer's toolkit. Google Cloud's latest DORA Report, an annual benchmark for software engineering practices, captures this seismic shift: 90% of technology professionals now use AI at work, up 14% from last year.
 
The numbers tell a story of transformation. Developers spend a median of two hours daily interacting with AI tools. Sixty-five percent rely on AI for at least half their workflow, with one in twelve reporting that their work is almost entirely AI-mediated. Eighty percent of respondents say their productivity has improved. Nearly 60% say their code quality has gone up.

And yet—trust lags behind adoption.

The Trust Paradox
Here lies the central paradox: developers know AI makes them faster, but they don't necessarily trust it. Only a quarter of respondents report a "lot" or "great deal" of trust in AI outputs. Thirty percent admit they trust AI "a little" or "not at all."

In practice, this doesn't stop them from using it. Instead, it reframes AI as a productivity amplifier rather than an autonomous coder. Teams are learning to "trust but verify," weaving AI-generated code into pipelines guarded by version control, rigorous testing, and human judgment.

The DORA team describes AI as a "mirror and multiplier." In high-functioning organizations, AI magnifies strengths, accelerating throughput. In fragmented ones, it exposes weaknesses, amplifying instability.

Throughput Up, Instability Too
The report measures software delivery performance across two key axes: throughput (lead time, deployment frequency, recovery from failure) and instability (change fail rates, rework). AI improves throughput—teams are shipping faster and more often—but it also introduces new instability.

The paradox of speed without reliability looms large. High-performing organizations, the data show, are those that invest not just in AI adoption, but also in the systems, practices, and cultural scaffolding that sustain the gains.

Seven Archetypes of the AI Era
To map this landscape, DORA identified seven team archetypes. At one end are "Harmonious High-Achievers" (20% of teams), who strike a balance between stability, low burnout, and high throughput. At the other end are teams facing "Foundational Challenges" (10%), stuck in process gaps, and suffering from high burnout.

Between these poles, the taxonomy spans "Stable and Methodical" (steady but cautious), "High Impact, Low Cadence" (fast but brittle), "Legacy Bottleneck" (hamstrung by outdated systems), and "Constrained by Process" (dragged down by inefficiency).

This clustering underscores the report's core message: AI isn't a magic bullet. It reflects the organizational soil in which it's planted.

The Blueprint: DORA AI Capabilities Model
This year, Google Cloud introduced the DORA AI Capabilities Model, a blueprint for harnessing the potential of AI. The model identifies seven organizational capabilities that separate the AI success stories from the cautionary tales:

• Clear AI stance: Organizations that define and communicate policies build trust and accelerate adoption.
• Healthy data ecosystems: High-quality, accessible data amplifies AI's value.
• AI-accessible internal data: AI tools tied into internal systems provide context-rich assistance.
• Strong version control: Frequent commits and rollbacks keep AI-generated code safe.
• Small-batch workflows: Iterative work reduces friction and improves performance.
• User-centric focus: Teams prioritizing UX see outsized benefits from AI adoption.
• Quality internal platforms: Robust internal developer platforms amplify gains but require careful governance.

The throughline: AI effectiveness is a systems problem. Adoption without scaffolding creates local wins but organizational drag. Adoption plus foundational practices drives systemic change.

Platforms, Pipelines, and Value Streams
Supporting this model is the broader industry shift toward platform engineering. Ninety percent of organizations now have at least one internal platform; three-quarters have dedicated platform teams. High-quality platforms smooth friction, improve morale, and enable scale—though, paradoxically, they also raise instability slightly as delivery accelerates.

Value stream management (VSM) has emerged as the counterbalance. By visualizing and measuring workflows, VSM ensures that productivity gains are aimed at the right problems, aligning AI acceleration with organizational strategy.

Augmenting vs. Evolving
The DORA report frames AI transformation in two modes: augmenting existing workflows or evolving new ones.

• Augmentation means patching bottlenecks—streamlining code review, modernizing data infrastructure, updating privacy protocols—so AI-driven productivity flows downstream.
• Evolution means designing AI-native workflows: delivery pipelines that dynamically test code, security systems that auto-detect threats, and collaboration models that blend human engineers with AI agents.

Most organizations will need to do both.

The Human Factor
Beyond productivity, the report tracks AI's sociocognitive effects. Developers using AI report higher "authentic pride" in their work, even if their perception of meaningfulness doesn't shift. Prompt engineering emerges as a distinct skill, transforming what it means to be an effective developer.

But risks remain. As AI tools enable experts to self-serve, junior developers risk being excluded from traditional skill-building pathways. The report suggests organizations must balance productivity with intentional skill development—measuring learning as carefully as throughput.

The New Baseline
A decade ago, DORA reports focused on elite performers and the practices that distinguished them. In 2025, the story is different. AI is the new baseline. What matters now isn't whether you use AI, but whether your organization has the foundations to use it well.

The closing note of the report is less a warning than a challenge: AI adoption is inevitable, but AI transformation is optional. The organizations that thrive will be those that treat it not as a tool, but as a reason to reimagine how software is built.