GCP – Unlock software delivery excellence and quality with Gemini Code Assist agents

According to DORA’s latest research – the Impact of Generative AI in Software Development report – AI tools are making software developers feel more productive, focused, and satisfied. They’re even writing better code and documentation more quickly. But the research uncovered a paradox: these individual gains do not translate directly into improved system-level performance. In fact, DORA’s platform-agnostic research found that increased AI adoption is correlated with:

• Decreased software delivery throughput: A 25% increase in AI adoption is associated with a 1.5% reduction in delivery throughput.

• Lower software delivery stability: A 25% increase in AI adoption is associated with a 7.2% reduction in delivery stability.

This presents a puzzle. How can AI simultaneously boost individual productivity and negatively impact overall system performance?

According to the research, the key is this: more code isn’t always better. This doesn’t mean AI is inherently detrimental to software delivery. It highlights the importance of strategic AI adoption. Focusing solely on code generation, without addressing other constraints in the system, is like adding more lanes to a highway that ends in a single-lane tunnel. Traffic will simply pile up.

 In this post, we’ll explore how the research suggests organizations can unlock software delivery excellence.

Gemini Code Assist agents

Gemini Code Assist for GitHub, released on February 25 in a public preview as part of Gemini Code Assist for Individuals, meets developers in the pull request workflow. This code review agent brings AI-assistance to your code review process, enhancing code maintainability, reviewing code against team-specific best practices, and supporting continuous integration. This means developers have more benefits, such as:

1. Improved documentation quality

2. Fast feedback

3. Easy migration of code bases

Any developer hosting repositories on GitHub can sign-up and install the Gemini Code Assist for GitHub from the GitHub marketplace today and learn more about the agent in our documentation. 

Improve documentation quality

High-quality documentation is consistently linked in DORA research to improved team performance, organizational performance, and developer well-being. The agent helps improve documentation quality by:

• Automating documentation generation and updates.

• Providing instant answers to developer questions about the codebase.

• Facilitating knowledge sharing and accelerates onboarding.

• Reducing tech debt (as the agent documents, it improves understanding).

By automating this often-tedious task, developers can focus on higher-value work, improving flow and job satisfaction.

Provide fast feedback

Fast, quality feedback loops help build confidence and trust in new tools and processes. Feedback also helps individuals, teams, and organizations change tactics when they see that things aren’t working as expected. Improving test automation and test data management helps teams get feedback faster and “shift left” on quality by:

• Generating comprehensive test cases.

• Prioritizing tests based on risk assessment.

• Identifying and flagging flaky tests.

• Reducing the reliance on manual testing.

• Increasing code coverage

Automated testing can reduce change failure rates and improve overall software reliability.

Migrate code bases

Keeping applications updated is required to maintain security, reliability, and maintainability. However, upgrades and migrations can be risky, time consuming, and difficult to prioritize. The agent can streamline the transition from old systems and dependencies by:

• Rewriting code from one language, framework, or version to another.

• Providing a full set of test cases to ensure the new implementation is correct.

• Offering comprehensive review to catch any errors that might have been missed.

• Delivering a set of documentation for the new implementation, saving you time and effort.

Use Gemini Code Assist agents to reduce technical debt 

Now that we have introduced you to some ways Gemini Code Assist agents can impact your SDLC, let’s look at a specific example every software team faces: Managing technical debt. 

We’ve built Gemini Code Assist to be more than just a coding companion—it’s a strategic partner in conquering technical debt and building a foundation for sustainable software delivery.

Technical debt isn’t about bad code, per se. It’s about the consequences of prioritizing speed over perfect design or taking shortcuts to meet immediate needs. It can be the old library that no one has time to update, the duplicated code that’s faster to copy-paste than refactor, the “temporary” workaround that becomes permanent, or the documentation that never gets written. It can arise from a lack of understanding, poor training, or even the unexpected retirement of a tool or service. It manifests in many forms. In fact, researchers from Google identified ten key categories.

The impact of this accumulated debt is profound. DORA’s research consistently demonstrates a strong correlation between technical debt and decreased software delivery performance. Technical debt doesn’t just slow down development; it actively undermines an organization’s ability to innovate, adapt, and compete.

Strategies for paying down technical debt

Just as financial debt requires a deliberate repayment plan, technical debt needs a strategic approach to remediation. There’s no magic bullet, but DORA’s research and industry practices point to several key strategies:

1. Prioritize and visualize: The first step is making the debt visible. Use tools like code analysis platforms, dependency trackers, and documentation coverage reports to quantify the problem. Value Stream Mapping (VSM) is invaluable for identifying bottlenecks and areas of waste caused by technical debt.

2. Embrace continuous improvement: Technical debt reduction should be an ongoing process, not a one-time project. Integrate it into your daily work.

3. Empower teams: Give teams the autonomy and resources they need to address technical debt. 

4. Modernize teams and architecture: Migrate to loosely coupled teams and architecture to reduce dependencies and make it easier to isolate and address technical debt in specific components.

5. Establish clear change management practices. Reduce bottlenecks and wait times associated with change by following a streamlined, clear, and collaborative change process, focused on reviews rather than blocking.

These strategies, when implemented consistently, create a virtuous cycle. Reducing technical debt frees up developer time and energy, allowing them to focus on building new features and further improving the system.

Gemini Code Assist:  Your ally in the fight against technical debt

By focusing on the entire SDLC, not just code generation, and by providing specialized capabilities for tasks like code review, documentation, and testing, Gemini Code Assist helps ensure that AI-generated code is thoroughly vetted and integrated responsibly. This reduces the likelihood of introducing errors and accumulating technical debt, leading to higher quality and more maintainable code.

You can customize the behavior of Gemini Code Assist to align with your organization’s specific coding environments, standards, preferred styles, and established best practices. Google Cloud is committed to a continuous improvement approach, guided by DORA’s research. We encourage our customers to:

1. Identify bottlenecks: Use DORA’s four key metrics and other performance indicators to pinpoint the constraints in your SDLC.

2. Plan targeted improvements: Select specific DORA capabilities to enhance, leveraging Gemini AI agents where appropriate.

3. Experiment and measure: Implement changes and rigorously track their impact on your performance metrics.

4. Adapt and iterate: Based on the data, refine your approach, continuously seeking improvement.

This iterative, data-driven approach, combined with the power of Gemini Code Assist agents, allows organizations to move beyond the hype and realize the true potential of AI in software development.

Try Gemini Code Assist today

The future of software development is a partnership between human ingenuity and artificial intelligence. Gemini AI agents are designed to be that partner.

• Try Gemini Code Assist today. Experience how our AI agents can transform your software delivery process.

• Explore Gemini Code Assist further with our documentation. See the product’s capabilities.

• Read the DORA research to dive deeper into the science behind high-performing teams.

Don’t just write code faster. Build better software, deliver it more reliably, and empower your teams to achieve their full potential.