Clarice AI founder runs company with ten AI agents
Benjamin Eckert operates Clarice AI with zero human engineers, relying on ten AI agents for product development. He argues that the CTO role is shifting from management to hands-on agent orchestration and specification-driven verification.
Benjamin Eckert, founder of Clarice AI, operates his company with zero human engineers, relying entirely on ten AI agents to handle product development. He rejects the term 'vibe-coding' in favor of a specification-driven development process where agents must define changes, write development plans, break them into tasks, implement, review, perform security reviews, and conduct static and live tests. Eckert personally reviews product specifications but not individual lines of code, arguing that verification should focus on functional and non-functional requirements rather than code aesthetics. He mandates that all software changes pass the OVALS Application Security Verification Standard, a 300-point criteria catalog, and uses Test-Driven Development to prevent regressions.
Eckert argues that traditional team sizes of four to eight engineers are obsolete; he predicts a shift to one-person pods or 'Product Engineers' who combine product and engineering roles. He states that the only relevant KPI for AI transformation is how many people are needed to ship the same output as before, noting that many companies still allocate only 200 euros for agentic engineering budgets, rendering them ineffective.
Eckert, who progressed from Backend Engineer at mobile.de to Director of Engineering, Acting CTO at Bubble, and Interim CTO at Benefix, observes that the CTO role has drastically changed in 2026 due to generative AI. He notes that private equity and venture capital investors now seek CTOs who are hands-on and capable of orchestrating agents, rather than pure managers. Clarice AI also utilizes AI to synthesize customer behavior and predict user needs, offering a complement to traditional user research. The platform allows companies to simulate customer reactions to product decisions within minutes, which is particularly effective in B2C contexts where simulated personas are robust. In B2B environments, the tool provides directional insights rather than definitive answers, with the company explicitly advising clients on the validity of the synthetic data.