Artificial intelligence is rapidly reshaping how software is built and delivered. Development teams everywhere are experimenting with AI coding assistants, chatbots, and automation tools. But simply adding AI into the workflow is not a guarantee of faster or better outcomes. The real question is how to harness AI effectively so that it truly boosts productivity and quality. To answer this, Google’s DORA research team, known for defining industry-standard DevOps performance metrics, recently studied what separates organizations that thrive with AI from those that struggle. In their 2025 research, DORA introduced the AI Capabilities Model, highlighting seven key practices that amplify the benefits of AI in software development. These seven capabilities provide a roadmap for companies eager to turn AI investments into tangible improvements in delivery and performance. In this blog, we will break down each of DORA’s seven AI capabilities and show how you can apply them using Atlassian’s suite of tools. From planning and code to collaboration and governance, Atlassian’s ecosystem is well-suited to help teams embed these best practices and get the most out of AI. Let’s explore what these capabilities are and how you can start implementing them today.

What is the DORA AI Capabilities Model?

The DORA AI Capabilities Model is a framework emerging from DORA’s latest State of AI-Assisted Software Development report in 2025. It identifies specific technical and cultural practices that enable development teams to succeed with AI. In other words, it’s not enough to simply introduce AI; you also need the right conditions in place for AI to make a positive impact. Through extensive research and data, DORA found that teams who excel with AI tend to have certain fundamentals in common, from clear policies to strong engineering habits. By focusing on these fundamentals, organizations can avoid common pitfalls and instead see AI drive real improvements in speed, quality, and team satisfaction.

Seven Key Capabilities for AI-Driven Development Success

Clear and Communicated AI Stance: A well-defined, well-communicated organizational position on how AI tools should be used. Teams need explicit guidelines on where AI assistance is encouraged or forbidden, how to handle AI-generated code or content, and what approval or oversight is required. Clarity here reduces confusion and builds trust in using AI responsibly.
Healthy Data Ecosystems: High-quality, unified, and easily accessible internal data. This means your documentation, code, and process data are organized and maintained so that both humans and AI can draw on reliable information. A healthy data ecosystem ensures that AI recommendations are based on accurate, up-to-date knowledge from within your organization.
AI-Accessible Internal Data: Connecting AI tools to your internal knowledge sources. When AI has access to relevant project wikis, repositories, tickets, and runbooks, it can provide context-specific help rather than generic suggestions. Integrating AI with internal data turns it into a more powerful assistant tailored to your company’s domain.
Strong Version Control Practices: Mature version control habits to manage the increased volume and velocity of AI-generated changes. This includes frequent commits, thorough code reviews, and the ability to easily revert or roll back changes. Robust version control processes prevent chaos as developers, whether human or AI, rapidly iterate on code.
Working in Small Batches: An iterative development approach where work is delivered in small, frequent increments. Small batch work, a staple of DevOps excellence, becomes even more vital with AI in the mix. It ensures that AI-generated outputs are integrated and tested continuously, which amplifies the positive impact of AI while minimizing the risk of compounding errors.
User-Centric Focus: A strong emphasis on end-user experience and feedback throughout development. When teams measure success by user outcomes, AI efforts stay aligned with real needs. A user-centric focus means AI is used to solve actual customer problems, and teams validate that AI-driven changes are truly beneficial to users.
Quality Internal Platforms: Reliable internal platforms and “golden path” tooling that streamline development. This refers to having well-supported internal systems for building, testing, and deploying software, such as CI/CD pipelines, developer portals, and standardized environments. Quality internal platforms let teams leverage AI at scale because common tasks are automated and development friction is low.

With these seven capabilities in mind, the next step is figuring out how to nurture them within your own organization. The good news is that if you’re using Atlassian products like Jira, Confluence, Bitbucket, and others, you already have a strong foundation to build upon. Let’s delve into each capability and explore practical ways to implement it using the Atlassian stack.

Software team leveraging AI tools and data for development.

Applying DORA’s AI Capabilities with Atlassian Tools

Now that we’ve covered what the DORA AI Capabilities Model entails, how can you put these ideas into practice? Below, we go through each of the seven capabilities and outline concrete steps and tips for implementing them using Atlassian’s ecosystem. These recommendations will help ensure your team not only adopts AI, but does so in a sustainable, effective way.

Clear and Communicated AI Stance

To establish a clear AI stance, start by creating a brief, explicit policy on AI usage within your projects. Document where and how AI tools are permitted or prohibited:

Define AI usage guidelines: Write a policy page in Confluence that outlines acceptable uses of AI, such as code generation, documentation suggestions, or test case generation, and any areas off-limits, such as sensitive code or customer data.
Communicate and reinforce: Make sure every team member knows the policy. You can reference this AI policy page in onboarding materials and even link it in Jira project templates or Confluence space homepages for easy visibility.
Integrate approval steps: In Jira workflows, consider adding a lightweight approval or review step for changes that involve AI-generated work. For instance, if a developer uses an AI tool to create a large piece of code, a Jira automation rule could assign a senior reviewer to double-check it. This ensures accountability without heavy bureaucracy.
Govern responsibly: Include any data handling rules in your policy. Teams should understand what internal data, if any, they can feed into AI tools. By having these rules clearly published, you reduce uncertainty and help everyone feel more confident about experimenting with AI within safe boundaries.

By openly publishing and discussing your AI stance, you set a tone of responsible innovation. Developers get clarity on expectations, which removes hesitation and prevents misuse. Atlassian tools make it easy to broadcast and embed these guidelines directly into the team’s daily workflow.

Healthy Data Ecosystems

A healthy data ecosystem means treating your internal documentation and knowledge as a first-class product. Atlassian offers the tools to cultivate this:

Single source of truth: Use Confluence as the central hub for documentation. All project decisions, designs, and how-to guides should reside in organized Confluence spaces rather than scattered in emails or personal notes.
Ownership and freshness: Assign clear ownership for important pages or knowledge areas. Atlassian’s Compass can help map out components or services and link to their docs, with designated owners for each. Establish a routine, perhaps on a quarterly schedule, where content owners update Confluence pages to prevent stale information.
Consistent labeling and archiving: Leverage labels in Confluence to tag content by project, topic, or status. This metadata helps both people and AI find the right documents quickly. Set up archiving rules or reminders for outdated pages so that irrelevant information doesn’t pollute your knowledge base.
Documentation as part of done: Encourage a culture where updating documentation is part of completing any major task. Jira workflows can include a step to remind assignees to link or update Confluence pages when closing out an epic or story. This keeps your data ecosystem continuously healthy.

Investing effort in documentation and knowledge upkeep pays dividends when AI tools come into play. The better your internal data, the better context AI has to assist your team. Atlassian’s integrated nature, with Jira issues linking to Confluence pages and other assets, makes it feasible to maintain this ecosystem without extra overhead.

Team members organizing and updating documents on a knowledge base.

AI-Accessible Internal Data

Once your internal knowledge base is in good shape, make sure your AI tools can tap into it. Atlassian’s AI assistant Rovo thrives when it has access to rich internal data. Here are some ways to make that happen:

Integrate AI with project spaces: Connect Atlassian’s Rovo or similar AI assistants to the content sources that matter. For example, link Rovo to specific Confluence spaces, Jira projects, and Bitbucket repositories. With proper access control in place, the AI can then index and draw from these sources when answering questions or generating suggestions for your team.
Prioritize high-value knowledge: Start by feeding AI the most critical and high-value data. This might include your Confluence spaces that contain architecture decision records, troubleshooting runbooks, FAQs, and design docs. By curating the input, you ensure the AI gives answers based on trusted, important information.
Use role-based permissions: Maintain your existing permission schemes so that the AI only accesses information that a given user is allowed to see. Atlassian’s permission settings will extend to its AI features as well. This way, a marketing user asking Rovo a question won’t accidentally surface an engineering-only page, for instance.
Continuous learning: As your team creates new pages or code, periodically update what the AI can access. The Atlassian platform can automatically include new Jira issues or Confluence pages in its index, but it’s wise to have someone oversee that the AI is staying in sync with your latest knowledge. This might involve re-indexing or manually prompting the AI to learn from a new key document when it’s added.

By making internal data AI-accessible in a governed manner, you transform your AI assistant from a generic help bot into a knowledgeable team member. Atlassian’s integrated tools plus AI can then answer complex questions like “Where is the design spec for project X?” or draft relevant test cases based on your own past incident reports, adding real value to your development workflow.

Strong Version Control Practices

When AI is contributing code changes or suggestions at a rapid pace, disciplined version control becomes even more important. Atlassian’s Bitbucket, or whichever Git-based platform your team uses, should be configured to enforce best practices:

Trunk-based development: Aim to keep a simple branch strategy. Many high-performing teams use a trunk-based approach, essentially keeping a single main branch with very short-lived feature branches. This minimizes merge conflicts even when AI is generating code suggestions for multiple developers.
Frequent commits: Encourage developers, and by extension their AI assistants, to commit early and often. Frequent commits with descriptive messages help track changes that AI tools contribute. It’s easier to pinpoint if an AI-introduced change caused an issue when commits are bite-sized and specific.
Mandatory pull reviews: Use Bitbucket’s branch protection rules to require pull requests for all merges into main branches. Every AI-written code snippet should still go through the normal peer review process. In Bitbucket, you can also integrate automated checks such as linting, tests, or even AI-powered code review comments so that quality issues are caught early.
Easy reverts and rollbacks: Set up your CI/CD pipelines to make reversions simple. Bitbucket Pipelines or your deployment tool should allow one-click rollbacks to a previous version. Practice using these rollback procedures in drills. By normalizing the process of reverting a deployment, the team will be more confident trying AI-assisted changes, knowing that if something goes wrong, it’s quick to undo.

Strong version control acts as a safety net and accountability system in an AI-accelerated development environment. Atlassian’s tools support these practices out of the box, letting you harness AI while still maintaining control over your codebase.

Developers collaborating on code with a Git repository on screen.

Working in Small Batches

Even with AI speeding up certain tasks, the wisdom of working in small batches still holds true. In practice, teams should break down work into bite-sized pieces and integrate changes frequently. Use Jira to split user stories into small tasks that can be completed in days, not weeks, so that any AI-generated code is delivered in a controlled scope. Employ feature flags to release new AI-assisted features rapidly but safely. This approach lets you deploy code into production behind toggles until it’s fully tested and ready for users. Developers should keep branches short-lived and rely on continuous integration using Bitbucket Pipelines to run tests on every commit, which helps catch issues early. It’s also crucial to limit work in progress in Jira; avoid launching too many parallel AI experiments at once, and finish what’s in progress before starting new tasks.

By working in small increments and shipping often, you make it much easier to evaluate the impact of AI contributions. If something goes wrong, you only have to inspect a small change set. If things go right, you deliver value to users sooner. Atlassian’s project tracking and CI tools help keep this rhythm steady and sustainable, even as AI accelerates parts of the development process.

User-Centric Focus

Even as your team adopts advanced AI tools, it is critical to keep a laser focus on the end user’s experience. Every AI-driven effort should tie back to a real customer need or improvement. To ensure this, start by capturing ideas and feedback in Jira Product Discovery so that each feature is linked to a clear user outcome. When planning sprints in Jira, define acceptance criteria in terms of user impact. For example, instead of a technical goal like “implement an AI recommendation engine,” frame it as a user benefit: “help users find relevant content faster with AI-driven recommendations.” This keeps the team oriented on delivering value rather than just building fancy algorithms.

Next, incorporate feedback loops at every stage. As you release AI-assisted features to a small beta group, collect input from users or stakeholders and document it in Confluence or Jira. Use that feedback to iteratively refine the product. Finally, measure success with user-centric metrics, whether it’s customer satisfaction scores, task completion times, or adoption rates of a new AI-powered feature. Track these metrics on Jira dashboards or Confluence status pages to visibly connect AI initiatives to user outcomes. If an AI-generated solution isn’t moving the needle on things that matter to users, take that as a signal to adjust your approach. The ultimate goal is a better experience for your customers, not just new technology for its own sake.

Quality Internal Platforms

The final capability is about the internal tools and infrastructure that developers rely on every day. High-performing teams invest in reliable “golden path” systems, which are streamlined, well-documented ways to build, test, and deploy software. Atlassian’s suite can serve as the backbone of these internal platforms. For example, you might standardize build and deployment pipelines using Bitbucket Pipelines or Bamboo, and document the steps in Confluence so everyone follows the same playbook. Atlassian Compass can act as a developer portal to catalog all your services, tech components, and best practices, making it easy for engineers and even AI agents to discover the approved tools and deployment patterns for each project.

It’s also important to offer self-service capabilities on this platform. Using Jira Service Management, you can set up automated workflows for common tasks. For instance, a developer could file a ticket to provision a new development environment or database, triggering an automated process to handle it. Meanwhile, treat your internal platform like a product in itself: monitor your CI/CD pipelines’ performance, keep your Jira and Confluence spaces organized and performant, and allocate time for platform maintenance. A robust, user-friendly internal platform means that when AI tools generate lots of output or suggestions, your underlying systems can handle the increased throughput. By smoothing out friction in the development process, quality internal platforms allow both human and AI contributors to deliver value faster and more consistently.

Wrapping Up

The era of AI-assisted software development is here, and the difference between merely using AI and truly benefitting from it comes down to these core practices. Google’s DORA research has distilled what successful teams do: they set clear guidelines, keep their data and code hygiene high, work iteratively, focus on users, and invest in solid tools and platforms. The exciting part is that none of these capabilities are out of reach. With the Atlassian stack at your fingertips and a thoughtful implementation, any team can start improving in each of these areas.

As an Atlassian Solution Partner, Atlas Bench has seen firsthand how adopting these seven capabilities can transform software delivery. Whether you’re drafting an AI usage policy in Confluence or revamping your CI pipeline in Bitbucket, the key is to start small and continuously improve. Pick one or two capability areas to focus on first, make tangible changes, and measure the impact. Over time, you’ll build a culture and system where AI truly accelerates your delivery of value to customers.

Interested in learning more? Contact Atlas Bench to book a complimentary consultation with our Certified Atlassian Experts.

undefined

Atlas Bench is an Atlassian Platinum Solution Partner with deep Cloud Specialization, helping Data Center customers move beyond migration and fully transform on the Atlassian Cloud Platform. We deliver Atlassian’s System of Work end to end through four collections: Strategy, Teamwork, Service, and Software, connecting planning to execution, aligning teams around outcomes, modernizing service delivery, and accelerating software development.

We are also recognized as an Atlassian World Class Software Development Partner of the Year (Americas), and we bring AI leadership to every engagement. By leveraging Rovo for automation and purpose built agents, we reduce manual work, unlock knowledge across tools, and speed up decision making so organizations realize measurable value faster in cloud.

Implementing DORA’s 7 AI Capabilities with Atlassian Tools