What It Really Takes To Make AI Agents Work

https://nfx.com/post/onboarding-ai-agents
  • 63% of leaders prioritize AI; 91% feel unprepared to act.
  • RPA = rigid rule-following; APA = RPA reliability + LLM adaptability.
    • APA recovers when conditions change; RPA breaks silently.
  • Accountability layer: AI must log each decision as verifiable code.
    • Maisa’s “Chain of Work” satisfies enterprise audit requirements.
  • Context layer: agents need unwritten org knowledge, not just explicit rules.
    • Like a great hire who absorbs culture, voice, and decision nuance.
  • Coordination layer: multi-agent orgs need inter-agent protocols + permissioning.
    • No winner-take-all; network effects accrue to platform layer.
  • B2A frontier: agents paying agents, signing contracts, using agent-native software.
  • Market thesis: agents turn labor into software; TAM described as trillions.

Hacker News

  • “Show HN: Libretto – Making AI browser automations deterministic” (134 pts, 57 comments)
  • “Show HN: Stage – Putting humans back in control of code review” (130 pts, 110 comments)
  • “Show HN: Runtime security for AI agents (injection, tool abuse, data exfiltration)” (1 pts, 1 comments)
  • “Show HN: An AI workflow to automate your LinkedIn job search” (2 pts, 0 comments)
  • “Show HN: I built a full LLM chat client as a Neovim filetype” (3 pts, 0 comments)

Anna Piñol, Partner at NFX · 2025-02-26 · Read on nfx.com


Type Link
Added Feb 26, 2025
Modified Apr 21, 2026