By Q2 2025, 100% of our 150 developers were actively using AI IDEs, and over 80% reported a productivity increase of more than 30%.
- Adoption: 100% usage across 25+ active projects.
- Measured impact: Over 80% report productivity gains above 30%. In small 0-to-1 projects, improvements go up to 50%.
- Next frontier: Our next step is developing AI Orchestrators, developers who don't just use AI IDEs, but skillfully combine tools, context, and intent to build faster, smarter, and with greater impact. It's not a destination we've fully reached yet, but a direction we’re actively exploring and investing in.
Since late 2024, Rootstrap Engineering committed to a strategic objective: "Shaping the Future Developer Role for an AI-Driven World." One fundamental aspect of this goal is the use of AI IDEs, and more specifically, using them well and to their full potential.
1. It's not about using AI IDEs. It's about using them right.
AI IDEs are now part of everyday development. According to Stack Overflow’s 2025 Developer Survey, 84% of developers are already using or planning to use AI tools in their workflows. But while usage keeps rising, confidence is starting to drop. Positive sentiment declined for the first time, from over 70% in 2023 and 2024 to just 60% in 2025. Even more telling: 46% of developers say they actively distrust the accuracy of AI tools, compared to only 33% who trust them. Just 3% report “highly trusting” the output.
This shift makes one thing clear: most teams have already adopted AI, but many are still figuring out how to get real value from it. That’s where Rootstrap has focused from the beginning, not just on adoption, but on effectiveness.
At Rootstrap, we made it our goal to use AI IDEs with intent and focus. Our objectives were clear:
- Accelerate delivery while also improving quality, always tied to the client’s business goals. In some projects, quality means stronger automated tests and code consistency. In others, it's faster iterations, better maintainability, or minimizing regressions in production. AI helps us move faster and build smarter, based on what quality means for each client and context.
- Improve the developer experience by minimizing repetitive, low-value tasks and enabling engineers to focus on high-impact work, like connecting code to product and business outcomes.
- Equip our engineers to lead the transition into a new era of software development.
But simply using an AI IDE doesn’t guarantee performance. The real gains come from how teams approach adoption. At Rootstrap, it started with clear leadership, a shared vision, and specific objectives. We empowered technical multipliers to lead by example and invested in our internal communities to foster collaboration and learning. These elements created the foundation, a culture that encourages experimentation, open discussion, and consistent iteration, which we’ll explore in more detail later on.
From that base, we discovered that two elements were particularly transformative. The first were Rules: structured workflows that guide the AI assistant to follow project-specific patterns and constraints, ensuring consistency and quality across codebases. The second were MCPs (Model Context Protocol), which integrate tools like JIRA, Figma, Databases, Playwright, or GitHub directly into the coding workflow. By bringing this context into the IDE, engineers can move faster while staying aligned with product requirements, design assets, and data sources.
By Q2, adoption of these practices was already significant. Half of the team had used Rules, and close to 80% of them reported finding them useful or very useful. MCPs were also gaining traction, with around 40% of developers experimenting with integrations, Figma, JIRA, and database tools ranking as the most valuable. To accelerate their use, we leaned on our internal technical communities and open-sourced our recommended setups in a public GitHub repository. Demos, and training sessions became part of our onboarding playbook, ensuring that every engineer had both the knowledge and the practical tools to make the most of AI IDEs from day one.
2. How we measured: quarterly surveys + delivery data
Since Q4 2024, we’ve run quarterly internal surveys to track AI IDE usage, perceived impact, and tool preferences. With response rates consistently above 80%, these surveys provide a representative view of how our engineering team experiences and applies AI tools.
To validate these insights, we paired the survey data with project-level delivery metrics, ensuring that what developers reported aligned with what we observed in actual project outcomes. Together with Project Managers and Engineering Managers, we tracked indicators like velocity, cycle time, sprint completion ratio, merge time, and other signals of delivery efficiency. This combined approach helped us cross-check perceived productivity gains against real improvements in how work moved through our pipelines.
By the end of Q2 2025, the numbers spoke clearly:
- 100% of the engineering team was using AI IDEs regularly.
- Over 80% reported productivity increases greater than 30%, while keeping an A score in maintainability across 100% of our projects.
- The most used tools were Cursor (76.5%), Copilot (14.7%), and Windsurf (6%).
- 47.2% had adopted Rules, with nearly 80% rating them useful or very useful.
- 40.3% had tried MCPs, with Figma, JIRA, and database tools as the top integrations.
Beyond metrics, the surveys also included broader dimensions, such as developer experience, onboarding ease, and collaboration impact, and invited qualitative feedback. These insights helped us identify patterns, understand blockers, and adapt strategies to further encourage usage across teams.
While we frequently discuss tool usage, LLM configurations, and best practices in our technical communities, each developer is free to choose not only the AI IDE and model that best fit their needs, but also the broader toolchain that suits their workflow. This includes the combination of MCPs most relevant to their project context, or even other AI-powered tools like terminals or automation agents. The ecosystem evolves rapidly, and what works best today might not be the best option tomorrow.
3. Lessons by project type and client maturity
As we expanded AI IDE adoption across more than 30 active projects, we learned that impact varies significantly depending on project type and client maturity. The numbers are only the starting point, behind them are lessons on where AI creates the most leverage and where its role is evolving fastest.
In small MVPs and rapid prototyping contexts, especially when the product scope and value proposition are well defined, efficiency gains reach up to 50%. Even though AI models are increasingly capable of handling broader context, the benefits are most visible when projects start from scratch or are still in their early stages. A large share of effort in these situations goes into building common application features such as sign-up flows, authentication, or CRUD operations, tasks where AI assistance saves substantial time and accelerates validation.
In larger 0-to-1 projects, focused on building scalable foundations, we consistently saw around 30% improvements. This is where the change from previous years is most striking. AI tools have become far better at understanding the context of more advanced projects, assisting developers not only with repetitive work but also with new feature development. Although engineers must still assume greater responsibility for architecture and business-critical design, the AI provides meaningful support that reduces friction and increases velocity.
In mature products with established architectures, the efficiency gains were smaller, averaging around 20%, but still highly valuable. These projects require managing vast amounts of historical context, which remains a challenge for AI. Nonetheless, value emerges by isolating components, applying AI to documentation, automated testing, or large-scale refactors. In these cases, maintaining shared Rules across teams becomes essential, ensuring consistency and enabling AI to operate effectively even in complex environments.
Finally, in very early-stage projects where the focus is simply on prototyping or shipping small MVPs, there are scenarios where traditional AI IDEs are not the only option. We also evaluate tools like Replit, Bolt, or Lovable, which enable “vibe coding” approaches. Depending on the complexity of the MVP, these platforms can accelerate delivery further and even allow designers or product managers to contribute directly to early builds.
Another important insight is that most of the gains developers experience today are still centered around the act of coding. These benefits are more easily realized in well-scoped, well-defined projects where AI assistants can operate with greater clarity. However, the real bottleneck in many initiatives is not the coding itself, but the upstream challenge of building the right product. Engineers often play a key role in this process as well, shaping requirements, aligning with stakeholders, and iterating on feedback. As Andrew Ng recently pointed out, product definition and management are becoming the new bottlenecks in the AI-driven development lifecycle source. Recognizing this is essential to fully unlocking the next wave of productivity.
4. How we drove adoption and effective use
Tooling alone doesn’t drive change, people and processes do. At Rootstrap, we treated AI IDE adoption as a change management challenge, not just a technical rollout. That meant combining leadership alignment, peer influence, and community-driven experimentation.
We began with a clear engineering vision centered on "Shaping the Future Developer Role for an AI-Driven World", our strategic goal for the engineering department. This vision was consistently reinforced in leadership meetings and team spaces. To make it tangible, we relied on technical multipliers: engineers who showcased practical use cases inside their projects and helped others onboard faster. Their influence proved far more effective than any top-down mandate. While leaders actively promoted adoption, it was the peer-led examples that truly resonated, making change feel both real and attainable.
Our technical communities became the backbone of adoption. These forums allowed engineers to refine Rules, test MCP integrations, and openly share wins and failures. The key was that developers could see peers within their same stack and context getting value from AI, while also learning from technical specialists who led by example. That combination, seeing someone with a similar background and respected experts in the organization both benefiting from the tools, proved far more effective than being told to adopt AI by someone outside their domain.
We also focused on early detractors, not ignoring them but addressing their blockers directly. Leaders played an active role in helping them overcome concerns, whether it was fear that AI would generate low-quality or insecure code, or anxiety about losing control over their own work. We reinforced that engineers remained fully accountable for the output, and that AI was there to augment, not replace, their expertise. At Rootstrap, developers value building high-quality software, and we’ve learned that “quality” isn’t a fixed standard, it depends on each client's business goals, product maturity, and constraints. Part of our approach, then, was to clearly explain how AI IDEs can be allies in achieving those goals, by improving test coverage, reducing regressions, catching security issues early, or streamlining code reviews, all while giving engineers more time and headspace to focus on what really matters in each context.
Finally, leadership involvement was critical. By consistently tying AI IDE usage back to business value, faster delivery, better quality, happier clients, leaders made it clear that adoption was not an experiment, but a strategic priority. They also reinforced that AI is a critical tool for us, one that enables developers to become more effective, more productive, and more focused on business and product outcomes. Our vision is clear: AI will empower developers, not replace them.
What we learned is that no single action was enough. Success came from combining top-down vision with bottom-up experimentation, showing developers both the why and the how. That blend made adoption stick and turned AI IDEs into a natural part of our culture.
5. What’s next: developers as AI orchestrators
“Shaping the Future Developer Role for an AI-Driven World” means anticipating how the role will evolve in the short, medium, and long term. The next step is moving from developers simply using AI tools to developers orchestrating them. This is no longer about applying MCPs or enforcing Rules, those practices are already standard. Instead, the AI Orchestrator is someone who coordinates several AI systems simultaneously, guiding them to work on different tickets or tasks, aligning their outputs, and ensuring coherence with the product and business vision.
This shift represents a major evolution compared to how the role was once understood. There was a time when each line of code was considered a craft in itself. But history shows how the developer’s scope has continually expanded: from writing in ASCII to higher-level languages like C and Java, from Ruby and Python to modern frameworks and libraries. Each step moved developers further away from low-level execution and closer to a broader perspective where developers think about how each piece fits into the bigger picture. AI now takes this transformation to a new level, positioning developers not only as technical experts, but as orchestrators of multi-agent systems that can autonomously plan, execute, and deliver.
Adoption and good usage were just the beginning. Scaling Rules, experimenting and adopting MCPs, and exploring background agents are all part of this evolving journey. But the shift doesn't end there. What comes next is the ability to orchestrate multiple AI agents assigning tasks, supervising their progress, and integrating outputs into coherent results.
We've already begun testing tools that enable this level of orchestration. From Cursor’s background agents and Claude Code Flow to platforms like Conductor, TaskMaster, and Magnet, we’re exploring how AI agents can go beyond passive assistance to take on more autonomous, parallel problem-solving roles. These are early steps, but they mark a clear direction: one where developers act as orchestrators of multiple intelligent components, and our role as leaders is to actively guide this transition, not just react to it.
Becoming an AI Orchestrator also requires new skills and cultural shifts. At Rootstrap, we see critical competencies as: critical thinking and problem-solving, mastery of AI techniques such as RAG and multi-agent systems, continuous learning and adaptability, curiosity, strong product and business orientation, communication and storytelling, and effective collaboration. These skills are not abstract; we embed them into our hiring practices, onboarding programs, and performance evaluations, so that every developer grows into this expanded role.
Technology alone will not define the future developer. What matters is the ability to orchestrate AI, connect technical work to business value, and lead change with curiosity and intent. That’s the journey we are on, and we see it as the defining evolution of our craft.
6. Conclusion: real productivity, clear strategy
As mentioned earlier, it's increasingly rare to find developers who don’t use AI IDEs. But not everyone uses them effectively, if there's even a clear definition of what 'effective' means. At Rootstrap, we’ve made this a strategic focus. AI IDE adoption here is not just encouraged, it’s a deliberate, evolving initiative that combines:
- Rigorous measurement.
- Tailored adoption by project type.
- Community-led implementation.
- A clear vision of the future developer role.
For our clients, this translates to faster delivery, stronger quality, and a team that adapts fast to change. For our team, it means growing with a mindset that embraces change, leads with understanding, and stays ahead.
If you're leading a development team and want to unlock the full value of AI in software engineering, let's talk.
.webp)
