Section 7 of 7 — 100 use cases, 24 tools

Enterprise AI use cases

100 ways companies put AI to work across HR, Finance, Supply Chain, Procurement, Security, and IT Operations. Each use case links to published ROI evidence, measured adoption where it exists, and the models and tools that lead the work.

17 use cases in Human Resources. 12 rely on capabilities with controlled-trial ROI evidence, and 6 have measured adoption in the Anthropic Economic Index.

Tools in this directoryWorkday IlluminateSAP SuccessFactors (Joule)BambooHR

Already on SAP, Oracle, or Microsoft?

What your platform ships for Human Resources
SAP
SAP SuccessFactors with Joule
  • HR Service Agent · Answers employee questions about HR policies. Techzine, Oct 2025
  • People Intelligence Agent · Analyses workforce data and recommends retention and compensation actions. Techzine, Oct 2025
  • Career and Talent Development Agent · Supports managers in career conversations with their teams. Techzine, Oct 2025

Part of SAP Business AI; assistants roll out through 2026, and Autonomous HCM is planned for November 2026. Anthropic’s Claude is a primary reasoning model for Joule agents (announced May 2026).

Oracle
Oracle Fusion Cloud HCM
  • Workforce Operations · Speeds scheduling approvals and reduces payroll issues. Oracle, Mar 2026
  • Career Advancement Command Center · Promotes internal mobility by connecting employees to open roles. Oracle, Apr 2026
  • Contract Compliance Workspace · Automates employment contract execution and analysis. Oracle, Apr 2026
  • Hiring Workspace for Store Managers · Cuts recruiting admin and speeds hiring workflows. Oracle, Apr 2026

Agents and AI Agent Studio are included with Fusion Cloud subscriptions at no additional cost.

Microsoft
Copilot Cowork
All functions
  • Copilot Cowork · Takes a described task and runs it end to end across Word, Excel, PowerPoint, Outlook, Teams, and connected systems, returning finished work. Microsoft 365 Blog, Jun 2026

Cowork needs a Microsoft 365 Copilot licence plus usage-based Copilot Credits, and is off by default for admins to enable. Security Copilot is included with Microsoft 365 E5 and E7. Copilot Cowork runs on Anthropic models (Claude Opus 4.8 and Sonnet 4.6).

“Built into” marks use cases a named SAP, Oracle, or Microsoft agent does directly, per the vendor’s announcement. ROI evidence is per capability, from published studies. Measured adoption is each use case’s share of Claude.ai or API usage in the Anthropic Economic Index, shown only where an exact match exists. Model chips show the directory models rated highest for that kind of work on LMArena (CC-BY 4.0).

By model

Highest-ROI use cases for each model

No study measures ROI model by model, so this ranks each model’s suitable use cases by the strength of published evidence for the use case, then by the model’s independent rating for that kind of work on LMArena.

Claude Opus 5

5 enterprise use cases fit this model. They’re ranked by the strength of published ROI evidence, then by its independent rating for the work on LMArena.

  1. 01
    Research, analysis & reportingControlled trial
    Strategy, consulting, finance
    • ▸ Rated #1 of 22 directory models on LMArena text (hard prompts), the closest independent rating for this work
    • ▸ Built for this kind of work (Language Models)
  2. 02
    Marketing, content & salesControlled trial
    Marketing & sales
    • ▸ Rated #1 of 22 directory models on LMArena text (creative writing), the closest independent rating for this work
    • ▸ Built for this kind of work (Language Models)
  3. 03
    Document & data processingControlled trial
    Operations, legal, finance back office
    • ▸ Rated #1 of 9 directory models on LMArena document, the closest independent rating for this work
  4. 04
    Customer supportField study
    Customer service
    • ▸ Rated #1 of 22 directory models on LMArena text (overall), the closest independent rating for this work
    • ▸ Built for this kind of work (Language Models)
  5. 05
    Agentic workflow automationEmerging
    Cross-functional operations
    • ▸ Rated #2 of 6 directory models on LMArena agent, the closest independent rating for this work
The evidence

Where AI pays off

Measured returns from controlled trials and field studies, with the caveats the same research found. Every figure links to its source.

Research, analysis & reporting

Strategy, consulting, finance

Controlled trial
+40%higher-quality output, 25% faster, 12% more tasks done

758 Boston Consulting Group consultants · Dell’Acqua et al., Organization Science

Caveat: On a task beyond the model’s abilities, consultants using AI were 19 points less likely to get it right. Know where the frontier is. Dell’Acqua et al., Organization Science

Top-rated in this directory for this work
Claude Opus 5#1Gemini 3 Pro#2GPT-5.6#3
Ratings: LMArena text (hard prompts)

Marketing, content & sales

Marketing & sales

Controlled trial
−40%time on professional writing tasks, with 18% higher quality

453 college-educated professionals · Noy & Zhang, Science (2023)

Caveat: Human-only teams still made better images, and AI-assisted output was more homogeneous. Ju & Aral (2025)

Top-rated in this directory for this work
Claude Opus 5#1Gemini 3 Pro#2Gemini 3 Flash#3
Ratings: LMArena text (creative writing)

IT & security operations

IT, security

Controlled trial
+35%accuracy, with 30% less time per task

IT administrators in randomised trials of Security Copilot · Bono & Xu, Microsoft (2024)

Agentic workflow automation

Cross-functional operations

Emerging

Agents that plan and execute multi-step processes across tools: search, updates, hand-offs, and follow-ups.

Caveat: Little controlled evidence on end-to-end agents yet. Measure a baseline and keep a human checkpoint before scaling. Wang et al. (2026)

Top-rated in this directory for this work
GPT-5.6#1Claude Opus 5#2Claude Sonnet 5#3
Ratings: LMArena agent

Voice, transcription & localization

Customer service, marketing, HR

Emerging

Transcribing calls and meetings, voice agents and voiceover, and translating content for new markets.

Enterprise software

Where to look first, by function

Most enterprise software wraps a frontier model in workflows, data connections, and guardrails for one part of the business. These are the products to evaluate first in each function.

01
Function 1 of 7

Human Resources

Workday

Workday Illuminate

  • Unifies HR and Finance in one platform, with a skills graph and attrition-risk prediction
  • Best-in-class people analytics and strong global compliance coverage
  • Enterprise-only pricing and long, multi-month rollouts
  • Overkill for organizations under roughly 1,000 employees
SAP

SAP SuccessFactors (Joule)

  • Native S/4HANA integration keeps HR data in sync with finance and operations
  • Multi-country payroll support across 46+ countries
  • Complex interface that needs real training investment
  • Heavy reliance on consultants to implement well
BambooHR

BambooHR

  • Fastest implementation of any major HR platform, typically 2 to 4 weeks
  • Clean interface with high employee adoption and transparent per-employee pricing
  • Designed for small and mid-sized organisations; very large enterprises may outgrow it
  • Native payroll is US-only; other countries rely on partner integrations
02
Function 2 of 7

Finance & Accounting

SAP

SAP Joule

  • Purpose-built finance agents for cash management and invoice processing inside S/4HANA
  • Reports a 65–80% cut in manual touch for accounts-payable automation
  • Value is concentrated inside the SAP estate; weak outside it
  • Requires real SAP expertise to configure and maintain
Oracle

Oracle Fusion AI Agents

  • Built-in agents across Oracle's own financial close, procurement, and reporting modules
  • Strong fit for organizations already standardized on Oracle Cloud ERP
  • Same lock-in pattern as SAP; most value stays inside Oracle's ecosystem
  • Agents are included at no extra cost, but only for Fusion Cloud subscribers
BlackLine

BlackLine

  • Focused specialist for financial close, reconciliation, and controls automation
  • Plugs into almost any ERP rather than requiring one specific vendor
  • Narrower scope than a full finance suite; doesn't replace core ERP finance
  • Best value when paired with, not instead of, a broader finance platform
03
Function 3 of 7

Supply Chain & Procurement

o9 Solutions

o9 Solutions

  • Enterprise knowledge graph unifies structured and unstructured planning data across a full multi-tier network
  • A Gartner Magic Quadrant Leader, with early agentic demand-planning features
  • High implementation effort and cost given how configurable it is
  • Like any planning tool, it can’t capture planners’ undocumented judgment on its own
Blue Yonder

Blue Yonder

  • End-to-end suite from planning through warehouse execution, with deep retail and CPG capability
  • A 2026 Gartner Leader, now with a conversational agent layer across the platform
  • Its “cognitive” AI stack is newly rebuilt, so maturity is still being proven
  • Breadth of the suite makes implementation heavier than point solutions
Kinaxis

Kinaxis Maestro

  • Concurrent, end-to-end orchestration with governed agent workflows that “plan with you”
  • Gartner Leader for eleven consecutive years, strong in automotive, high-tech, and industrial sectors
  • Enterprise-scale pricing with no published cost tiers
  • Getting value requires carefully modelling your own planning rules and data
SAP

SAP Joule (Supply Chain & Procurement)

  • Named agents for bid analysis, supplier onboarding, production planning, and trade classification
  • Anthropic’s Claude is a primary reasoning model behind Joule agents, announced at Sapphire 2026
  • Built around SAP’s own applications and data, so non-SAP estates gain less
  • Many Joule Assistants roll out through 2026; check availability module by module
Oracle

Oracle Fusion SCM & Procurement Agents

  • Agents for planning, sourcing, maintenance, logistics, and order management inside Fusion Cloud SCM
  • Included with Fusion Cloud subscriptions at no additional cost, with AI Agent Studio to build more
  • Only relevant for organisations running Oracle Fusion Cloud
  • Agentic applications launched in 2026, so production track records are short
05
Function 5 of 7

Sales & Marketing

Salesforce

Salesforce Agentforce

  • Deep CRM and Data Cloud integration gives agents rich customer context for personalized outreach
  • Powerful workflow automation built on existing Salesforce Flows
  • Grounding agents in unstructured content needs Data Cloud, which adds cost
  • Most value depends on clean, well-structured Salesforce data
HubSpot

HubSpot Breeze

  • Built directly into HubSpot's inbound marketing and CRM suite, easy for SMB and mid-market teams
  • Ties content generation to the same data as email, forms, and pipeline
  • Most valuable only if your team is already standardized on HubSpot
  • Less powerful for long-form or highly specialized copy than dedicated writing tools
Jasper AI

Jasper

  • Strong at brand-voice-consistent marketing copy at campaign scale
  • Popular, mature tool with templates built around real marketing workflows
  • Per-seat cost adds up quickly for larger marketing teams
  • Output still needs human editing for accuracy, originality, and nuance
06
Function 6 of 7

Customer Service & Support

Intercom

Intercom Fin

  • Reports a 76% average resolution rate across roughly 12,000 customers, and climbing
  • Native AI-first helpdesk that support teams can manage without engineering help
  • Delivers the most value only once it's connected deeply into your business systems
  • Per-resolution pricing ($0.99 each) can be hard to forecast at high volume
Sierra

Sierra

  • Highly customizable agents built around a company's own policies, with strong enterprise governance
  • Well suited to complex, high-volume support operations
  • No public pricing; sold as premium enterprise contracts
  • Implementations are custom projects rather than self-serve setup
Zendesk

Zendesk AI

  • Simple to turn on for teams already running their helpdesk on Zendesk
  • Over 1,800 marketplace integrations to extend it
  • Built as an add-on to Zendesk’s help desk rather than as an AI-native agent
  • Capabilities are tied to staying inside the Zendesk ecosystem
07
Function 7 of 7

IT Operations & Security

ServiceNow

ServiceNow (Now Assist)

  • Extends into an AI Control Tower that discovers, governs, and secures AI agents across the enterprise
  • Strong workflow automation for IT service management and incident response
  • Most value is concentrated for organizations already running ServiceNow ITSM
  • Adds another layer of cost and complexity to an already complex deployment
Darktrace

Darktrace

  • Self-learning anomaly detection that spots novel threats without relying on known signatures
  • Can take autonomous containment action against active attacks in real time
  • Needs careful tuning to cut down on alert noise
  • Priced and positioned for enterprise-scale security budgets
Glean

Glean

  • Enterprise search across 275+ apps makes it a strong self-service base layer for IT knowledge
  • Near real-time permission enforcement keeps sensitive systems properly scoped
  • Needs real upfront integration work across many source systems
  • Overlaps with, rather than replaces, dedicated ITSM tools like ServiceNow
Microsoft

Microsoft Security Copilot

  • Included with Microsoft 365 E5 and E7 since November 2025, with no separate licence
  • Agents inside Defender, Entra, Intune, and Purview for phishing triage, access policy, and vulnerability fixes
  • Most valuable on a Microsoft security stack; mixed-vendor estates get less
  • Several agents launched in preview, so confirm general availability before planning around them