AI-powered development tools are fundamentally changing who can build software in the enterprise. Platform engineering teams now face a new challenge: enabling non-technical business users to safely build and deploy applications at scale. This shift - from platforms built exclusively for developers to platforms that serve developers, agents, and business users alike - represents one of the most significant evolutions in platform engineering's short history.

Main insights

• SIXT successfully deployed over 700 applications built by citizen developers (non-technical business users who build software using AI-assisted tools) in less than a year, addressing both high-fidelity requirements and long-tail business needs
• AI-generated code at SIXT jumped from 25-28% across 2025 to over 80% in Q1 2026, but individual productivity gains haven't yet translated to proportional team-level velocity increases
• Platform teams can enable safe citizen development through tool-agnostic platforms, automated guardrails, curated data access layers, and transparent infrastructure management
• The key to scaling citizen development lies in injecting best practices at code generation time and automating security, compliance, and deployment entirely

Kaspar von Grünberg led this roundtable discussion with Boyan Dimitrov, CTO of SIXT. Dimitrov oversees SIXT's engineering hubs across Munich, Bangalore, Kyiv, and Lisbon, bringing over 15 years of experience in cloud and distributed systems to the conversation about how platform teams can support this emerging user group.

You can watch the full discussion here if you missed it.

From better requirements to long-tail enablement

SIXT's journey toward citizen development began in Q2 2024, when the engineering team noticed a fundamental shift in how they worked. "We found ourselves writing way less code and spending most of our time in specs, spending most of our time basically asking AI to do something for us," Dimitrov explained. That observation sparked a direct question: what if the business teams closest to the work could build solutions themselves?

Two motivations drove the initiative. First, SIXT wanted higher-fidelity requirements, faster. Rather than requirements changing hands multiple times, domain experts could iterate on working prototypes before involving engineering teams. "If we allow this to happen, we can get way high fidelity prototypes and much better requirements way quicker," Dimitrov noted.

Second, the platform team recognized an opportunity to address the long tail of deprioritized projects - the small-scope work that, despite significant technology investment, engineering teams would simply never reach. Individually minor, these projects collectively represent significant efficiency gains when business users can build them independently.

By Q3 2024, SIXT had already onboarded early users and realized the potential extended far beyond initial expectations. The platform now supports over 700 applications built by citizen developers across HR, sales, and various internal teams.

What citizen developers are actually building

The applications emerging from SIXT's platform span a surprisingly wide range. "We have anything to do with very ad hoc analysis or dashboards insights people never had before," Dimitrov explained. The ability to go custom rather than one-size-fits-all has unlocked deep analysis of small process segments that would never have justified dedicated engineering resources.

Beyond dashboards, users are building full automations, process workflows, and integrations into systems of record. Product teams create working prototypes and even run A/B tests before formal engineering involvement. HR and sales teams - not traditionally tech-savvy groups - are active builders on the platform.

Governance boundaries remain important. SIXT allows reading from systems of record but stays conservative about write operations. "We are pretty conservative when it comes to writing into systems of record where we would say you know this is where maybe you want to involve an engineering team," Dimitrov explained. This balance between empowerment and risk management is deliberate.

The architecture: Tool-agnostic platforms with automated guardrails

SIXT made a critical early design decision to remain tool-agnostic. Rather than standardizing on a single AI coding assistant, the platform supports Cursor, Claude, Codex, and other tools through a common integration layer. "We wanted to allow our citizen developers to use whatever tool they could be comfortable with," Dimitrov said.

The platform injects SIXT context and best practices at code generation time through cursor rules, Claude skills, and MCP servers. This includes security requirements, scalability patterns, and access to existing platform building blocks. "We try to optimize around this already from the get-go and that means injecting a lot of SIXT context," Dimitrov explained.

A platform agent manages the entire deployment pipeline, hidden from citizen developers entirely. Once requirements are met, the platform agent builds and ships applications automatically - packaging anything, provisioning infrastructure, and injecting non-functional requirements like authentication, authorization, and monitoring without user input. "We had to throw basically everything we did in terms of CI/CD for our engineering platform," Dimitrov admitted. The new system represents a complete rethinking of delivery pipelines for a non-technical audience.

Securing enterprise data access at scale

Safe data access is one of the most critical challenges in enabling citizen developers at enterprise scale. SIXT solved this through a dedicated data access layer that acts as a single control point. "It understands our data. It understands the criticality behind that data," Dimitrov explained.

The platform exposes curated datasets rather than everything, and applications inherit the access levels and roles of the users who build them. This provides a "single choke point to control and govern" while maintaining appropriate security boundaries.

The citizen development platform also operates in isolation from other SIXT systems. "We assume that any of those applications can misbehave in many different ways and therefore we have to contain that blast," Dimitrov said. This containment ensures a broken application cannot impact production environments or other systems - a non-negotiable design principle at enterprise scale.

The productivity paradox: Individual gains vs. team velocity

The data on AI-generated code at SIXT is striking. AI-generated code rose from 25-28% across all of 2025 to over 80% in Q1 2026. Yet Dimitrov identified a fascinating disconnect between individual and team productivity. "Even though the individual engineers are more productive, I don't see that correlation yet." Cycle times haven't compressed proportionally, and teams aren't shipping 5-8x more despite individual engineers reporting 5-10x productivity gains.

The bottleneck is organizational, not technical. "At the speed at which you're then developing, you're restricted by the lines of communication," Kaspar observed. The coordination overhead between humans - alignment, handoffs, approvals - creates a flow problem. One delay cascades through the entire delivery chain.

Boyan believes solving this requires rethinking team composition and reducing human-to-human handover points. "Per capita, the amount of stuff you're taking on needs to be much larger to reduce the handover point between actual humans," he suggested. For platform teams, this is the next frontier: not faster code generation, but fewer coordination bottlenecks.

Scaling support without scaling headcount

As adoption grew, SIXT faced increasing support requests and onboarding load. The platform team's response was instructive: rather than iterating on documentation or creating training videos, they infused the onboarding experience directly into the AI tools users already work in.

"We go and we infuse that experience straight into skills, rules, whatever for Claude, for Codex, for Cursor where people work," Dimitrov explained. "People don't have to ask because magic things just happen." This approach eliminated most basic support requests entirely.

The remaining questions focus on scaling successful applications - "Hey, this thing is working for 10 people, how do I make it work for a thousand?" - which represent the natural graduation point where engineering teams take over ownership.

When engineering must take over

SIXT maintains clear boundaries around what citizen developers can own. Anything touching end-to-end customer experience requires engineering ownership. And for technically complex domains, the limits are explicit. When asked about building distributed systems handling millions of transactions per second, Dimitrov was direct: "Today that doesn't work. That simply doesn't work."

The models are "very convincing when they speak about concurrency and those types of problems, but they fundamentally don't understand them." At scale, the symbiotic integration between hardware, architecture, and software requires engineering expertise that AI-assisted citizen development cannot yet replicate.

The future: Platforms for developers, agents, and business users

The implications extend far beyond individual productivity. "Enterprise AI adoption without platform engineering" isn't viable, von Grünberg argued. As every vertical in an organization needs agentic capabilities, platform engineering teams become the providers of the substrate - context management, vector databases, RAG pipelines, and more.

Dimitrov sees the layer of responsibility between hyperscaler offerings and application code shrinking dramatically - "probably two, three, four times less than it is today." The task for platform teams is building experiences that work through natural language rather than requiring technical knowledge. That's a fundamentally different product design challenge than what most platform teams have tackled before.

​

If you enjoyed this, find here more great insights and events from our Platform Engineering Community.

​

If you want to dive deeper, explore our instructor-led Platform Engineering Certified Professional course and connect with peers from large-scale enterprises who are driving platform engineering initiatives.

​

Key takeaways

​

• Define governance boundaries before you scale. SIXT's success stems from establishing clear rules upfront - what citizen developers can build, which data they can access, and when engineering must take over. Starting with these guardrails prevents security issues and scope creep while enabling rapid experimentation.
• Inject best practices at code generation time. By embedding security requirements, architectural patterns, and platform context directly into AI coding assistants through rules and skills, you can enforce quality without manual oversight. This approach scales far better than traditional code review processes.
• Build tool-agnostic platforms that abstract infrastructure completely. Users must be able to work in whatever AI coding tool they prefer while your platform handles packaging, deployment, infrastructure provisioning, and non-functional requirements automatically. This future-proofs your investment as the tooling landscape continues to evolve rapidly.
• Address the organizational bottleneck, not just the technical one. Individual productivity gains from AI won't translate to team velocity until you reduce human-to-human handover points and rethink team composition. The communication protocol between humans - not code generation speed - is now the primary constraint on delivery.