Why 88-95% of enterprise AI agent pilots never ship — and what working teams do differently
ankitsharma112 · reddit · 2026-09-07
A practitioner with years of enterprise agent experience argues "ready for production" is the wrong frame: agents are ready only for a specific bounded job inside guardrails. Key points:
- The oft-quoted stat is 88% (some sources 95%) of agent pilots never ship — and it's almost never because the model is dumb
- The real killer is non-determinism: agents fail quietly and confidently in ways test suites never anticipated; 1/3 of catalogued production incidents trace back to unhandled tool failures that golden-path tests can't catch
- Teams that ship treat evals/observability as infrastructure, use deterministic guardrails and modular design (LLM only for intent, exact operations in typed testable code), gate high-stakes actions with human approval, and decompose early
- What works now: support triage, SDR/outbound, internal ops, coding assistance; banking/insurance lead while healthcare and public sector lag — the bottleneck is governance and data, not intelligence
More from coding & agent
- Dev ships two iOS apps and an Obsidian plugin in one go with Codex — vista8 · 2026-09-07
- Anthropic's Fermat project: agents failed until a shared coordination system fixed them — VraserX · 2026-09-07
- Korea's top game studio opens its animation/sound assets: VARCO 3D MCP and VARCO Sound — arrakis_ai · 2026-09-07
- Spotify cut Claude Code token usage by 90% by routing big file reads to cheap models — aliscodes · 2026-09-07
- Knowledge work is much harder for agents to automate than code, argues Matt Pocock — mattpocockuk · 2026-09-07
- Production-aware coding agents: closing the feedback loop after AI code generation — WasteAcanthaceae4938 · 2026-09-07