Of the 8 most-cited enterprise agentic AI vendor claims made in Q1 2026 (Salesforce Agentforce, Microsoft Copilot Agent Mode, Google Gemini Enterprise, Anthropic Claude for Enterprise, OpenAI Agents SDK, ServiceNow AI Agents, Workday Illuminate, SAP Joule), a minority remain Holding at 90-day review, a majority sit at Partial with at least one falsified component, and customer-cited ROI claims hold materially better than vendor-cited ROI claims — meaning the citation-source of an enterprise AI claim is a stronger predictor of its 90-day durability than the size of the vendor making it.

Claim created at publish. Review on 90-day cadence from 12 May 2026 (next due 10 Aug 2026). The citation-source pattern (customer-cited vs. vendor-cited durability gap) is marked source:'our-estimate' throughout the article — it is an editorial observation derived from cross-referencing 8 vendor claim sets against third-party analyst coverage and public customer case-study disclosures, not a measured metric from a named research study. The 8 Q1 2026 vendor claims being graded are those documented in the Q1 2026 article at /agentic-ai-got-real-q1-2026/ plus the specific product announcements cited in this scorecard. Anchor sources: Salesforce Q1 FY27 earnings call 21 May 2026 (Agentforce 3.0 GA), Microsoft Q3 FY2026 earnings 30 Apr 2026 (Copilot agent seat growth), Google Cloud Next 2026 9 Apr 2026 (Gemini Enterprise multi-agent preview), Anthropic Claude for Enterprise GA blog 8 Apr 2026, OpenAI Agents SDK 1.0 release notes 15 Apr 2026, ServiceNow Q1 2026 earnings 23 Apr 2026, Workday Illuminate GA press release 14 Apr 2026, SAP Joule roadmap update Apr 2026. Sister claims: AM-013 (Q1 2026 governance threshold article — the 8 claims being graded are downstream of that piece), AM-148 (GPT-5.5 vs. Opus 4.7 split-verdict — the model-routing read informs which vendor infrastructure claims are plausible). Trigger conditions to revisit before next cadence: (a) a named vendor re-states a graded claim as 'Not holding' in an earnings call or investor day with a named timeline revision — the scorecard row moves and the citation-source meta-pattern may strengthen or weaken; (b) Gartner or Forrester publishes a Q3 2026 Magic Quadrant update that materially reorders the Holding/Partial/Not-holding distribution; (c) a customer-cited ROI claim in the scorecard is retracted or materially revised by the citing customer — would be the first data point running against the citation-source durability thesis; (d) any of the 8 vendors ships a GA milestone that closes a graded capability gap within the review window, which would move a row from Partial to Holding.