"Who's liable when AI errs?" A delivery engineer on real enterprise AI blockers
2C_ornot2C · x · 2026-11-03
A reflective long thread from frontline enterprise AI delivery work. The author once thought enterprise adoption meant automating repetitive work with models, Agents, RAG and Workflow and proving efficiency in a demo — but real projects revealed owners worry about ROI, business teams fear workflow disruption, IT fears data leaks, and management worries about accountability. Even cutting a two-hour task to twenty minutes won't win AI control over approvals, customer data or business decisions.
The core pain: "When AI errs, who's responsible?" Completing tasks autonomously and bearing responsibility are entirely different things — hallucinations, API timeouts, wrong document retrieval, duplicate retries are all real risks to those who own business outcomes.
Enterprise workflows are also far from standardized: data scattered across Excel, internal systems and personal machines; undocumented expertise everywhere. The author's conclusion: what makes enterprises dare to adopt a system is the unsexy stuff — requirements analysis, data governance, permission design, logging, evals, human review, retry and exception handling. Model intelligence alone never settles a cost, organizational, accountability and habit problem.
More from AGI Musings
- Burkov: P=NP will never be proven unless a human writes the proof — burkov · 2026-11-03
- A green-text thread on p(doom): solar's 80% price drop undercuts AI doomerism — prasanna_says · 2026-11-03
- signull: Human EQ is rare because its inference cost is brutally expensive — signulll · 2026-11-03
- Software is shifting from winner-take-all to a restaurant-style long-tail market — tinyfool · 2026-11-03
- The Russell Conjugations of Intelligence: Why 'Just Matrix Multiplication' Offends Us for Brains but Not for AI — esrtweet · 2026-11-03
- AI safety researcher: assuming all emergence is risk may be the wrong frame — repligate · 2026-11-03