Felmo's AI-Driven Engineering and Cost Optimization
Felmo's CTO details the strategic shift from Cursor to Claude Code to manage API costs while maintaining high output. The analysis covers the reduction of human-in-the-loop processes, the rise of AI-generated code, and the evolution of developer roles in a tech-enabled veterinary startup.
Strategic Shift in AI Engineering Costs
Felmo, a tech-enabled mobile veterinary service, has restructured its software development workflow to address the rapid escalation of AI tool costs. The CTO highlights a critical pivot from pay-per-token API models, such as those used in Cursor, to flat-rate subscription services like Claude Code. This decision was driven by the fact that API costs exceeded annual budgets within weeks, whereas subscriptions offer predictable expenses and higher token quotas. This cost optimization is essential for startups that cannot absorb volatile infrastructure costs while scaling AI-assisted development.
Impact on Code Volume and Review Processes
The adoption of agentic coding tools has resulted in approximately 50% of code pushes being AI-generated, with line-of-code metrics indicating an even higher proportion. This volume surge has fundamentally altered code review practices. Traditional line-by-line peer reviews are being replaced by high-level architectural checks and AI-assisted linting. The team has moved beyond the "human-in-the-loop" model for routine tasks, relying on automated testing and AI-driven QA to validate changes. This shift allows for faster deployment cycles and reduces the cognitive load on developers, who can focus on complex problem-solving rather than syntax verification.
Evolving Developer Roles and Market Standards
As AI handles routine coding and testing, the role of the software engineer is expanding. Developers are now expected to manage specifications, collaborate closely with designers, and monitor business outcomes. This convergence of technical and product responsibilities is creating a new class of "superhuman" engineers who can iterate rapidly from idea to production. Furthermore, the widespread adoption of AI is raising the baseline quality of software across the market. Average standards for code quality, SEO, and app functionality are increasing, forcing companies to compete on customer value and market fit rather than basic technical execution. For Felmo, this means leveraging AI not just for efficiency, but to enhance the core service delivery for their veterinary clients, ensuring that the technology enables the business model rather than just supporting it.
Conclusion
The integration of AI into Felmo's engineering stack demonstrates a mature approach to cost management and workflow optimization. By prioritizing subscription models, automating quality gates, and expanding developer responsibilities, the company is positioning itself to maintain agility and competitiveness in a rapidly evolving technological landscape. This case study offers a blueprint for other startups seeking to harness AI without succumbing to cost overruns or process bottlenecks.
Key insights
-
Flat-rate AI subscriptions are more cost-effective than API-based models for high-volume coding tasks. This shift allows startups to predict expenses and scale usage without budget constraints.
Impact: Reduces infrastructure costs and prevents budget overruns, enabling sustainable scaling of AI-assisted development.
-
AI-generated code now constitutes a significant portion of production software, necessitating a shift from detailed peer reviews to high-level architectural oversight. This change reduces review bottlenecks and accelerates deployment.
Impact: Increases development velocity and allows teams to manage higher code volumes without proportional increases in review time.
-
The role of the software engineer is expanding to include product specification, design collaboration, and business outcome monitoring. This convergence creates more versatile and impactful team members.
Impact: Enhances team agility and ensures that technical solutions are aligned with business goals and customer needs.
-
AI adoption is raising the baseline quality of software across the market, making basic technical execution a commodity. Companies must differentiate through customer value and market fit.
Impact: Forces businesses to focus on unique value propositions and customer experience rather than relying on technical superiority alone.
-
Automated testing and AI-driven QA are replacing manual verification for routine tasks, reducing the need for human-in-the-loop processes. This shift improves efficiency and reduces cognitive load on developers.
Impact: Accelerates release cycles and allows developers to focus on complex problem-solving rather than routine testing.
Action items
-
Evaluate and switch from pay-per-token AI APIs to flat-rate subscription models to control costs. This involves analyzing current usage patterns and comparing the total cost of ownership for different tools.
Impact: Prevents budget overruns and provides predictable expenses for AI tooling, enabling better financial planning.
-
Implement AI-assisted code reviews and automated linting to reduce the burden of manual peer reviews. Focus on high-level architectural checks for AI-generated code.
Impact: Increases development velocity and allows teams to manage higher code volumes without proportional increases in review time.
-
Expand developer responsibilities to include product specification, design collaboration, and business outcome monitoring. Provide training and tools to support this broader role.
Impact: Enhances team agility and ensures that technical solutions are aligned with business goals and customer needs.
-
Automate quality assurance processes using AI-driven testing tools to reduce the need for manual verification. Focus on non-critical features first to build confidence in the system.
Impact: Accelerates release cycles and allows developers to focus on complex problem-solving rather than routine testing.
-
Differentiate your product by focusing on customer value and market fit rather than basic technical execution. Use AI to enhance the customer experience and service delivery.
Impact: Positions the business to compete effectively in a market where basic technical quality is becoming a commodity.
Quotes
“Wir sind kein Tech-Start-up, aber wir sind Tech-Enabled, die Tierärzte.”
“Ich würde behaupten, dass die Hälfte unseres Codes, also die, sagen wir mal, die Hälfte unserer Pushes AI-generiert sind.”
“Der Busfaktor ist definitiv ein viel geringeres Thema als noch vor zwei Jahren.”