S45 Cofounder Limits Hiring to Roles AI Cannot Do

Inc42 reports that S45 cofounder Aman Singh said the startup hires only for roles that cannot be performed by artificial intelligence. Inc42 reports that S45's workforce is largely split between engineers developing AI agents and what the outlet describes as "AI agent bankers" who operate those agents. Inc42 also reports comments from INDmoney that it is too early to set a fixed AI budget and that AI spend is unlikely to exceed 10% of tech expenditure. Inc42 additionally reports Open CEO Anish Achuthan saying AI usage will keep rising while falling token prices are reducing per-token costs.

Inc42's coverage frames S45's model around agent-based automation and operator roles rather than traditional front-office headcount. The article does not provide specific model names, architectures, or vendor relationships, nor does it include verbatim technical specifications or benchmark numbers.

Companies that adopt agent-driven workflows often concentrate hiring on engineering talent to build, monitor, and integrate agents, while shifting routine operational tasks to supervised agent operators. This pattern increases the importance of production ML engineering, monitoring, orchestration, and prompt/agent design skills across the stack.

Observers tracking early AI-native startups have reported similar trade-offs between human headcount and automation, and comments about keeping AI budgets modest reflect broader uncertainty around model inference economics. Inc42's reporting that INDmoney expects AI spend to remain under 10% of tech budgets aligns with public discussions about volatile token pricing and changing cost trajectories.

For practitioners and observers: monitor hiring postings and org charts for more startups to see whether roles labeled as "agent operators," prompt engineers, or ML reliability engineers appear at scale; track vendor disclosures on inference costs and effective token pricing; and watch for published post-deployment metrics (latency, failure modes, human-in-the-loop rates) that clarify the operational burden of agent-based services.