AGENTIC·CASEBOOKENTERPRISE AGENTIC AI USE CASE CATALOG
RUN #1 · 2026-07-04
Google Cloud AI GTM Intelligence · Daily Research Run

Where enterprise agentic AI is actually real — evidence-graded, field-ready.

A living catalog of agentic AI use cases across five industries, built from primary-source evidence — company announcements, earnings calls, and production metrics. Every entry carries a status, a confidence score, named adopters, and a Google Cloud play. Vendor marketing is a lead, not evidence.

14Use cases
10Deployed
2Pilot
2Emerging
5Industries

AExecutive Summary

Inaugural run, 2026-07-04. The evidence is no longer demos — it is operating metrics from named enterprises.

Signals of the run

  • 40,000 prior-auth decisions/day at a major US insurer — agents auto-approve the routine majority; clinicians see only judgment calls.
  • Deutsche Telekom × Google Cloud: RAN Guardian has taken 100+ autonomous network remediations; major events now handled in ~1 minute, not hours.
  • Hiscox: specialty quote turnaround compressed from 3 days to ~3 minutes (−99.4%).
  • Citi "Arc": 1.5M automated code reviews ≈ 100K engineer-hours returned; banks now buy agent platforms, not point agents.
  • Morgan Stanley is opening ShareWorks/Equity Edge to its clients' external AI agents — the first "agent front door" in FSI.

Google Cloud plays

  • Gemini Enterprise for CX — 7 named retail adopters in 6 months (Macy's live in 4 weeks; Home Depot, Ulta, Kroger, Lowe's, Papa Johns, Woolworths 🌏).
  • DT MINDR reference architecture — take it to every Tier-1/Tier-2 operator; Verizon-class RFPs are forming now.
  • Universal Commerce Protocol (NRF 2026, with Target/Shopify/Etsy/Wayfair; Walmart aligning) — Google as the rails of agentic commerce.
  • Healthcare wedge — prior-auth & denials agents on Vertex AI + Document AI + Healthcare API (FHIR), with a CMS regulatory forcing function.
  • Above-the-ERP layer in manufacturing — SAP Joule owns in-ERP steps; the cross-system orchestration layer is open field.

Patterns that matter commercially

  • Document-heavy decisioning agents (prior auth, underwriting, appeals) show 10–1000× cycle-time collapses — the fastest, best-quantified ROI anywhere.
  • Closed-loop operations agents went production in <12 months; telecom leads all industries (48% agentic adoption vs 26% cross-industry).
  • Platform-not-project buying: Citi Arc, Epic Agent Factory, ServiceNow Autonomous Workforce, Morgan Stanley super agents.
  • Voice is the production wedge in QSR and telco care — clearest labor math, highest containment.
  • The agent-to-agent economy is forming: enterprises treating other people's agents as a customer channel.

Competitor watch

  • OpenAI — Workspace Agents for enterprises; Burger King voice pilot; ServiceNow frontier-model deal. Now a platform rival, not just a model vendor.
  • Microsoft — Agent 365 ↔ ServiceNow AI Control Tower; won Uniper's SAP-procurement agents with Celonis. Owns the "governed agent estate" narrative in IT.
  • AWS — selling Rufus-derived shopping AI to retailers; Bio Discovery (Apr 2026) attacks pharma R&D budgets.
  • SAP / Epic / ServiceNow / Oracle — vertical software owners embedding agents where workflows live; hyperscalers must win the cross-system layer.
  • 🌏 JAPAC: Woolworths on Gemini CX; Hitachi & NEC agent launches; SoftBank sovereign-AI push shapes Japan deals.

BUse Case Catalog

Click any card for the full field brief — autonomous actions, human checkpoints, evidence with sources, metrics, Google Cloud mapping, and verbatim discovery questions.

CCommercial Potential

Weighted score: enterprise spend 35% · urgency 25% · repeatability 25% · Google Cloud fit 15%.

#IDUse caseSpendUrgencyRepeatGC fitScore

DReadiness

Scored on technology maturity, evidence status, typical data availability, integration complexity (inverted), and governance clarity.

▲ SELL TODAY

  1. Voice ordering (RET-QSR-001) — mature stack, bounded domain, Wendy's is the Google reference.
  2. Prior auth (HLS-PROV-PA-009) — deployed at scale, FHIR standards exist, CMS tailwind.
  3. Agentic CX (RET-CX-002) — productized; Macy's live in 4 weeks.
  4. Denials & appeals (HLS-RCM-010) — document-centric; approval loop simplifies governance.
  5. UW triage (FSI-INS-UW-006) — human-in-loop keeps model risk tractable.
  6. Subscriber care (TEL-CS-013) — established pattern, clear metrics.
  7. MRO requisitions (MFG-PROC-007) — bounded ERP actions with approval gates.

EMonthly Industry Heatmap — July 2026

Agentic activity by industry × business function, rated on evidence volume and confidence trend. Baseline month — movement tracking starts August.

IndustryCustomer service / CXSales & commerceOps / network / SCFinance / claims / rev cycleR&D / engineeringRisk & compliance
Retail & ConsumerHOTHOTWARMINGQUIETWARMINGQUIET
Financial ServicesWARMINGWARMINGQUIETHOTHOTWARMING
Manufacturing & SCQUIETQUIETHOTQUIETWARMINGQUIET
Healthcare & LSWARMINGQUIETQUIETHOTHOTWARMING
TelecommunicationsHOTQUIETHOTQUIETWARMINGQUIET

FField Action Kit

Conversation starters, objection handling, and a first-meeting talk track per industry — grounded in this run's evidence, liftable verbatim.

GEvidence Method

Every card carries exactly one status. The categories are never blurred; vendor marketing alone never supports DEPLOYED.

DEPLOYED

Running in production at a named company at meaningful scale — adopter-side or credible secondary source, dated.

PILOT

Named pilot, POC, or limited rollout with explicit pilot framing.

EMERGING

Repeatable pattern across ≥2 independent signals; no named production deployment yet.

SPECULATIVE

Plausible but unproven. Excluded from "ready now" rankings.

CONFIDENCE 1–5

5 = multiple named production deployments with quantified outcomes · 3 = named pilots or GA product with early customers · 1 = single promotional source.

AGENTIC BAR

The system must plan, use tools, or take multi-step autonomous action. Pure content generation or single-turn Q&A is excluded (e.g., Epic "Art" documentation is held at watchlist).