---
id: platforms-enterprise
related:
  - platforms-consumer-developer
  - consolidation-enterprise
  - architecture-trends
  - ai-financial-services
key_findings:
  - "Microsoft Copilot has 15M paid users but only 3.3% conversion rate — massive reach, weak monetization"
  - "Salesforce Agentforce 29K deals signals the agent-first enterprise shift is real, not theoretical"
  - "ServiceNow 80B workflows/yr is the largest operational AI training dataset in enterprise software"
  - "Oracle 1000+ agents leveraging data gravity moat — bundling strategy working"
---

# Enterprise AI Platforms Registry
**As of March 22, 2026 | Compiled from primary vendor sources and verified market data**

---

## Table of Contents
1. [Enterprise Suite Vendors](#enterprise-suite-vendors)
   - Microsoft 365 Copilot + Copilot Studio
   - Google Workspace AI (Gemini)
   - Salesforce Agentforce
   - ServiceNow AI (Autonomous Workforce)
   - Oracle AI (Fusion Cloud)
   - SAP Joule
   - Workday AI
2. [Dedicated Agent / Automation Platforms](#dedicated-agent--automation-platforms)
   - Kore.ai
   - UiPath
   - Zapier AI / Make
   - n8n
3. [Infrastructure / Model Providers](#infrastructure--model-providers)
   - OpenAI API
   - Anthropic API
   - Google Vertex AI
   - AWS Bedrock
   - Azure OpenAI Service
   - OpenRouter
   - Together AI / Groq / Fireworks

---

## Enterprise Suite Vendors

### 1. Microsoft 365 Copilot + Copilot Studio

**Current AI Agent Capabilities**
- Agent Mode launched in Word, Excel, and PowerPoint (December 2025): multi-step autonomous document creation and data analysis without user steering each step
- Agents can coordinate with each other on complex tasks by calling other agents as tools (multi-agent architecture, rolling out March 2026)
- Agent Mode in Office apps combines reasoning models for agentic productivity; Office Agent powered by Anthropic models for PowerPoint/Word creation in chat
- Copilot Studio enables low-code/no-code agent building; AI agents can be deployed in Microsoft 365 apps, Teams, Outlook, and OneDrive
- Microsoft Agent 365 serves as the control plane / governance layer for agents across Foundry and Copilot Studio
- Source: [Microsoft 365 Blog – Agent Mode Announcement](https://www.microsoft.com/en-us/microsoft-365/blog/2025/09/29/vibe-working-introducing-agent-mode-and-office-agent-in-microsoft-365-copilot/); [What's New February 2026](https://techcommunity.microsoft.com/blog/microsoft365copilotblog/what%E2%80%99s-new-in-microsoft-365-copilot--february-2026/4496489)

**Adoption / Scale**
- 15 million paid Microsoft 365 Copilot seats (reported January 2026 earnings call), up 160% year-over-year — but represents only ~3.3% of the 450 million commercial Microsoft 365 user base
- 90% of Fortune 500 companies have used or accessed Copilot (FY26 Q1 framing); 80%+ of Fortune 500 companies have active AI agents built with Copilot Studio or Microsoft Agent Builder (per Microsoft telemetry, November 2025)
- PwC: 200,000 seats deployed; Lloyds Banking Group reports 46 min/day saved per employee
- Recon Analytics (Jan 2026): Only 8% of users who tried Copilot alongside ChatGPT and Gemini preferred Copilot; paid subscriber market share fell from 18.8% to 11.5% between July 2025–January 2026 (-39%)
- Source: [The Register – 3.3% conversion rate, Feb 2026](https://www.theregister.com/2026/02/02/microsoft_ai_spend_copilot/); [Microsoft Security Blog – 80% Fortune 500](https://www.microsoft.com/en-us/security/blog/2026/02/10/80-of-fortune-500-use-active-ai-agents-observability-governance-and-security-shape-the-new-frontier/)

**Pricing Model**
- Microsoft 365 Copilot enterprise: **$30/user/month** (annual); promotional pricing ~$18 ran through early March 2026
- Copilot Business (≤300 users): **$21/user/month** (annual, or was promoted at $18)
- Copilot Chat: Free for all Microsoft Entra ID users with eligible Microsoft 365 subscriptions; agents priced on metered (pay-as-you-go) basis via Azure
- Copilot Studio: Sold as tenant-wide license at **$200/pack/month** (25,000 Copilot Credits per pack). Pay-as-you-go at **$0.01/credit**
- Billing rates: Classic answer = 1 msg; Generative answer = 2 msg; Agent action = 5 msg; Tenant graph grounding = 10 msg
- Source: [Microsoft Copilot Studio Pricing](https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/copilot-studio); [Enterprise Pricing](https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/enterprise); [Adoptify Licensing Guide](https://www.adoptify.ai/blogs/2026-microsoft-copilot-consulting-pricing-and-licensing-guide/)

**Key Differentiator**
- Deepest integration with existing Microsoft 365 workflows (Office apps, Teams, Outlook, SharePoint); largest enterprise user base globally with 430M+ M365 commercial seats; Copilot Studio enables rapid no-code agent building across the tenant

**Strategic Direction**
- Positioning Copilot as a "daily habit" / digital labor platform; evolving from copilot to multi-agent orchestration (agent-to-agent coordination); Microsoft Foundry as the enterprise agent development and deployment platform; tight NVIDIA integration (Nemotron models on Microsoft Foundry announced GTC 2026)
- Source: [Microsoft 365 Roadmap – Agent Mode PPT](https://www.microsoft.com/microsoft-365/roadmap?id=555889)

---

### 2. Google Workspace AI (Gemini)

**Current AI Agent Capabilities**
- Gemini AI assistant embedded across Gmail, Docs, Meet, Sheets, Slides, and Drive across Business Standard and above plans
- NotebookLM with expanded access (Business/Enterprise plans)
- Flow: visual AI workflow and agent builder (launched 2025); each AI Expanded Access user gets 2,000 monthly credits for Flow
- Workspace Studio: AI-powered automation builder (Rapid Release available, Scheduled Release mid-March 2026)
- Real-time speech translation in Meet (promotional access ongoing)
- Image and video generation via Nano Banana Pro and Veo 3.1 (requires AI Expanded Access from March 1, 2026)
- Gemini 3 Pro (most capable model) fully integrated for agentic workflows and multimodal tasks in Vertex AI / Workspace
- Source: [Google Workspace AI Expanded Access Help](https://knowledge.workspace.google.com/admin/gemini/ai-expanded-access); [Workspace Updates Blog – Feb 5, 2026](https://workspaceupdates.googleblog.com/2026/02/google-workspace-ai-expanded-access.html)

**Agent Count / Integrations**
- AI Agent Trends 2026 report (3,466 executives surveyed) outlines 5 major trends; agent builder ecosystem expanding through Google ADK (Agent Development Kit), Agent Engine, and Vertex AI Agent Builder
- 200K+ token context window on Gemini 2.5 Pro / Gemini 3 Pro; 1M token context on Gemini 2.5 Pro and select Gemini 3 models
- Source: [Google Cloud AI Agent Trends 2026 Report](https://cloud.google.com/resources/content/ai-agent-trends-2026)

**Pricing Model**
- Google Workspace tiers (annual, 16% discount):
  - Starter: **$5.60/user/month** (Gemini in Gmail only)
  - Standard: **$11.20/user/month** (Gemini in Gmail, Docs, Meet + more)
  - Plus: **$17.60/user/month**
  - Enterprise: Custom pricing
- **AI Expanded Access add-on** (available from Feb 5, 2026): Available to Business Standard/Plus and Enterprise Standard/Plus; unlocks higher limits on Workspace Studio automations, image/video gen, speech translation; pricing not publicly listed — requires admin purchase
- **AI Ultra Access add-on**: Highest limits for specialists; includes Gemini 2.5 Deep Think
- Google AI Pro (consumer): $19.99/month; includes Gemini 2.5 Pro, 1,000 AI credits/month
- Google AI Ultra (consumer): $124.99/3 months; 25,000 credits/month
- Source: [Google Workspace Pricing](https://workspace.google.com/pricing); [Workspace Help – AI Expanded Access](https://knowledge.workspace.google.com/admin/gemini/ai-expanded-access); [Finout Gemini Pricing 2026](https://www.finout.io/blog/gemini-pricing-in-2026)

**Enterprise Customer Count / Revenue**
- No specific enterprise customer count publicly disclosed for Workspace AI as of March 2026; Google Workspace serves 10M+ paying businesses globally (historical data)
- Google AI advertising integration under consideration (March 2026 reporting)
- Source: [MarketingProfs AI Update March 20, 2026](https://www.marketingprofs.com/opinions/2026/54448/ai-update-march-20-2026-ai-news-and-views-from-the-past-week)

**Key Differentiator**
- Native integration across Gmail, Drive, Docs, Meet ecosystem with no additional client software; Gemini 3 Pro's 1M+ token context window for deep document/research tasks; Google Search grounding for real-time web-aware agents

**Strategic Direction**
- AI Expanded Access model (tiered add-on introduced March 1, 2026) transitions from promotional to commercial; heavy investment in agentic workflow builder (Flow); MCP and A2A protocol support planned; potential advertising integration into Gemini assistant
- Source: [Google Workspace Updates Blog](https://workspaceupdates.googleblog.com/2026/02/google-workspace-ai-expanded-access.html)

---

### 3. Salesforce Agentforce

**Current AI Agent Capabilities**
- Agentforce 360 Platform: autonomous AI agents for Sales, Service, Marketing, Commerce, and IT workflows
- Atlas Reasoning Engine: configurable hybrid reasoning balancing LLM creativity with deterministic controls; supports RAG, retrieval, and multi-step planning
- Agentforce Builder: low-code/no-code builder for business users and admins; updated Feb 2026 with more admin-friendly UI
- Agents deployed in Slack, Einstein Trust Layer (data security); Agentforce in Slack saved 500,000+ employee hours in first year
- Engagement Agent: worked over 310,000 leads alongside sales team
- Einstein Trust Layer ensures zero data retention by LLM providers, PII masking, toxicity detection
- MCP server support for cross-agent orchestration with third-party systems
- Source: [Salesforce Agentforce 360 What's New](https://www.salesforce.com/agentforce/what-is-new/); [Salesforce FY26 Q4 Highlights](https://www.salesforce.com/news/stories/fy26-q4-highlights/)

**Deal Count / Revenue**
- **29,000 cumulative Agentforce deals closed** as of Q4 FY2026 (reported Feb 25, 2026), up 50% quarter-over-quarter
- **Agentforce ARR: ~$800 million**, up 169% year-over-year
- Agentforce + Data 360 combined ARR: **$2.9 billion** (includes $1.1B Informatica Cloud ARR), up 200%+ YoY
- 19–20 trillion AI tokens consumed; **2.4 billion Agentic Work Units (AWUs)** delivered to date, including 771M in Q4 FY2026
- 75% of Salesforce's top 100 Q4 deals included both Agentforce and Data 360
- >60% of Agentforce/Data360 bookings from existing customer expansions
- Deals >$1M up 26% YoY; deals >$10M up 33% YoY
- Source: [Salesforce Q4 FY2026 Earnings Release](https://investor.salesforce.com/news/news-details/2026/Salesforce-Delivers-Record-Fourth-Quarter-Fiscal-2026-Results/default.aspx); [Futurum Analysis](https://futurumgroup.com/insights/salesforce-q4-fy-2026-earnings-show-agentic-ai-scaling-guidance-steadies/)

**Pricing Model**
- Three concurrent models as of Feb 2026:
  1. **$2/conversation** (original Oct 2024 launch; still available but largely replaced)
  2. **Flex Credits: $0.10/action** ($0.15/Flex Credit; actions consume 120 credits each — ~$0.15 per action in most cases); launched May 2025
  3. **Per-user license: $125/user/month** (Agentforce digital labor framing, late 2025)
- Voice actions: 120 Flex Credits per 100 calls (~$0.18/100 calls)
- Agentforce 1 Edition and A4X (premium SKUs) "nearly tripled" QoQ bookings in Q4 FY2026
- Source: [Salesforce Agentforce Pricing Page](https://www.salesforce.com/agentforce/pricing/); [SaaStr – Pricing Evolution](https://www.saastr.com/salesforce-now-has-3-pricing-models-for-agentforce-and-maybe-right-now-thats-the-way-to-do-it/); [Monetizely – Pricing History](https://www.getmonetizely.com/blogs/the-doomed-evolution-of-salesforces-agentforce-pricing)

**Key Differentiator**
- Deep CRM data integration via Salesforce Data Cloud; Einstein Trust Layer for enterprise compliance; Atlas Reasoning Engine with hybrid LLM + deterministic planning; largest purpose-built agentic sales/service ecosystem

**Strategic Direction**
- Agentforce becoming Salesforce's primary growth engine targeting $63B revenue by FY2030 (11% CAGR); expanding Agentforce into IT Service Management (displacing ServiceNow customers); "digital labor" narrative with AWU metric; FY27 guidance of $45.8–46.2B with H2 organic reacceleration expected
- Source: [Constellation Research Q4 Analysis](https://www.constellationr.com/insights/news/salesforce-q4-strong-agentforce-picks-momentum); [The Silicon Review](https://thesiliconreview.com/2026/03/salesforce-record-q4-fiscal-2026-results-agentforce)

---

### 4. ServiceNow AI (Autonomous Workforce)

**Current AI Agent Capabilities**
- **Autonomous Workforce** launched Feb 26, 2026: orchestrates teams of AI specialists (not just individual task agents) with defined roles — Service Desk, Service Agent, Security Operations
- **Level 1 Service Desk AI Specialist**: diagnoses and resolves common IT issues (password resets, access provisioning, network troubleshooting) using enterprise knowledge bases, historical incident data, and proactive remediation workflows; 99% faster than human-handled cases (internal use)
- **EmployeeWorks**: combines Moveworks conversational AI + enterprise search with ServiceNow workflows; accessible via Teams and Slack; handles multi-system coordination with governance and audit trails
- **Agent Orchestrator**: central coordination layer managing AI specialists across complex workflows
- **AI Agent Studio** (Yokohama release): create agents, define agentic workflows, set triggers, test outcomes
- **Now Assist**: embedded GenAI across ITSM, CSM, HR, and security; includes incident summarization, chat summarization, resolution notes, email reply recommendations
- Now Platform processes **80 billion+ workflows per year**; LLM carries ~10% of lift; ServiceNow IP and 20 years of workflow data carries ~90%
- Build Agent: AI-powered app development on the Now Platform (Zurich release)
- Source: [ServiceNow Autonomous Workforce Press Release, Feb 26 2026](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-launches-Autonomous-Workforce-that-thinks-and-acts-adds-Moveworks-to-the-ServiceNow-AI-Platform/default.aspx); [Cloud Wars – Autonomous Workforce Detail](https://cloudwars.com/ai/servicenows-latest-ai-deliverables-automate-tasks-within-governed-workflows/); [ServiceNow + OpenAI Collaboration](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-and-OpenAI-collaborate-to-deepen-and-accelerate-enterprise-AI-outcomes/default.aspx)

**Scale / Revenue**
- ~1,000 AI customers and surpassing $200M in annual contract value for Pro Plus AI tier (end of 2024)
- Acquired Moveworks for **$2.85 billion** (completed December 2025), 20x+ Moveworks' 2024 revenue; strengthens conversational AI and enterprise search
- GTC 2026 demonstration of Autonomous Workforce governance (March 17, 2026)
- OpenAI named preferred intelligence capability for ServiceNow workflows; GPT-5.2 integration for frontier reasoning in enterprise workflows
- Source: [ServiceNow Moveworks Acquisition Close](https://www.moveworks.com/us/en/company/news/press-releases/servicenow-completes-acquisition-of-moveworks); [ServiceNow GTC 2026 Press Release](https://newsroom.servicenow.com/press-releases/details/2026/NVIDIA-GTC-2026-Governing-the-Autonomous-Workforce/default.aspx)

**Pricing Model**
- Subscription-based, quote-driven:
  - ITSM Standard: ~$100/fulfiller/month
  - ITSM Pro: ~$160+/fulfiller/month (AI features as add-on)
  - ITSM Pro Plus: ~$160+/fulfiller/month + 60% premium for Now Assist GenAI; Assist Packs for overages
  - Enterprise: custom; large enterprises typically $1M–$3M+/year in licensing alone
- Average contract: ~$130,080/year (varies significantly by module count)
- Now Assist features: consumption-based "Assist" tokens with monthly caps; Pro Plus tier required for GenAI
- Source: [NowTribe ITSM Pricing 2026](https://nowtribe.com/how-much-does-servicenow-itsm-cost-in-2026-a-complete-breakdown/); [Xurrent ServiceNow Pricing 2026](https://www.xurrent.com/blog/servicenow-pricing); [eesel.ai Pricing Breakdown](https://www.eesel.ai/blog/servicenow-pricing)

**Key Differentiator**
- 80B+ workflows/year providing verified "what good looks like" training data; outcome-guaranteed agents (not probabilistic black boxes); platform-native orchestration across ITSM, HR, Security, CSM in a single workflow graph

**Strategic Direction**
- Autonomous Workforce as the next frontier beyond copilots; Level 1 Service Desk AI Specialist targeting GA in Q2 2026; expanding from ITSM into adjacent workflows (HR, security, EX); Moveworks integration enabling conversational front-end to automated back-end; NVIDIA GTC 2026 governance framework demonstration
- Source: [ServiceNow GTC 2026](https://newsroom.servicenow.com/press-releases/details/2026/NVIDIA-GTC-2026-Governing-the-Autonomous-Workforce/default.aspx)

---

### 5. Oracle AI (Fusion Cloud)

**Current AI Agent Capabilities**
- **1,000+ AI agents** live inside Oracle Fusion Cloud Applications (horizontal back-office and industry-specific apps) — including hundreds of agents in the banking suite alone
- AI agents prebuilt and natively integrated in Fusion at **no additional cost** for Oracle Fusion subscribers
- Oracle AI Agent Studio for Fusion Applications: build, test, and deploy custom agents and agent teams enterprise-wide
- New CX agents (Feb 10, 2026): Marketing Campaign Agent, Sales Pipeline Agent, Service Resolution Agent for Oracle Fusion CX
- New SCM agents (Feb 10, 2026): Supply Chain Risk Agent, Service Parts Advisor Agent, Production Planning Agent
- ERP and HCM agents: cover finance, accounts payable, procurement, and HR lifecycle automation
- GTC 2026 showcases: "AI Investigator" agent for Financial Crime Compliance using Oracle AI Database + NVIDIA NeMo Agent Toolkit; hybrid model allocation (Nemotron-3 Nano vs. Super) per workflow step
- GPU-accelerated vector index generation in Oracle AI Database 26ai using NVIDIA hardware and cuVS (GA announced GTC 2026)
- OCI Supercluster with NVIDIA Vera Rubin platform (next-gen training and inference)
- NVIDIA Nemotron 3 Super model available via Oracle AI Database Model Import (first NVIDIA model in OCI Generative AI)
- Source: [Cloud Wars – Oracle 1,000 Agents, Mar 2026](https://cloudwars.com/ai/oracle-unleashes-1000-ai-agents-to-automate-entire-industries/); [Oracle CX AI Agents Announcement](https://www.oracle.com/news/announcement/oracle-ai-agents-help-marketing-sales-and-service-leaders-enhance-cx-2026-02-10/); [Oracle NVIDIA GTC 2026 Blog](https://blogs.oracle.com/cloud-infrastructure/oracle-nvidia-gtc-2026-key-announcements)

**Revenue / Scale**
- Oracle is among the largest enterprise software companies globally ($57B+ FY2025 revenue); cloud applications (including Fusion Cloud) account for a significant and growing share
- Fusion Cloud ERP, HCM, SCM, and CX serve tens of thousands of enterprise customers

**Pricing Model**
- Fusion Applications: subscription-based (per user or per module), enterprise contracts; AI agents included at no additional cost within existing Fusion subscriptions
- OCI Generative AI: token-based API pricing for hosted models via OCI
- Source: [Oracle Fusion AI Agents – No Additional Cost](https://www.oracle.com/news/announcement/oracle-ai-agents-help-supply-chain-efficiency-and-strengthen-resiliency-2026-02-10/)

**Key Differentiator**
- 1,000+ purpose-built vertical agents embedded in existing Fusion workflows at no added cost; Oracle AI Database 26ai provides secure, enterprise-grade data layer with GPU-accelerated vector search; banking suite with hundreds of domain-specific agents; deep NVIDIA partnership for model optimization

**Strategic Direction**
- Expanding AI agents across all Fusion modules; Open Agent Specification enabling multi-step, multi-model agent workflows; NVIDIA NeMo Agent Toolkit integration for step-level model optimization (quality vs. cost); sovereign AI cloud via Oracle Government Cloud; positioning OCI as the enterprise AI infrastructure platform
- Source: [Oracle NVIDIA GTC Blog – AI Database](https://blogs.oracle.com/database/oracle-ai-database-nvidia-collaboration-advances-enterprise-ai-at-nvidia-gtc-2026); [Oracle NVIDIA Key Announcements GTC](https://blogs.oracle.com/cloud-infrastructure/oracle-nvidia-gtc-2026-key-announcements)

---

### 6. SAP Joule

**Current AI Agent Capabilities**
- **Joule Studio**: generally available (GA) in Q1 2026; agent builder within SAP Business Technology Platform (BTP) enabling custom Joule agent creation with no-code and pro-code options
- **Joule Agent Builder**: GA as of Q1 2026; multi-agent coordination for cross-enterprise workflows
- **Joule Deep Research**: synthesizes internal SAP data and external intelligence for complex, multi-domain questions (beta/early adopter care, Q1 2026)
- **400+ AI-driven use cases** embedded across SAP applications (as of late 2025)
- **Agentic evolution (2026)**: Joule acts as cross-enterprise workflow coordinator, not just a Q&A copilot; proactively finds errors, manages tasks across departments, escalates only when human approval is required
- Role-based AI assistants in Joule: tailored to finance managers, HR professionals, procurement leads, etc.
- AI Agent Hub in SAP LeanIX: centralized governance dashboard for all AI agents
- Joule integrates with Microsoft 365 Copilot (shared skills and data across ecosystems)
- MCP server for SAP Commerce Cloud (planned Q2 2026); Agent-to-Agent (A2A) interoperability protocol supporting cross-vendor agent collaboration
- SAP-ABAP-1 foundation model trained on real-world ABAP code for code modernization/migration
- SAP predicts AI agents could support up to **80% of most-used business tasks** in SAP
- NVIDIA NIM microservices: up to 20% performance improvement vs. other serving engines; model-agnostic architecture allows BYO models via SAP AI Core
- Source: [SAP + NVIDIA GTC 2026 Partnership](https://news.sap.com/2026/03/how-sap-nvidia-advance-ai-enterprise-transformation/); [AIMultiple SAP AI Agents 2026](https://aimultiple.com/sap-ai-agents); [SAP Community Q1 2026 UX Update](https://community.sap.com/t5/technology-blog-posts-by-sap/sap-ux-q1-2026-update-part-1-ai-joule-sap-build-work-zone-sap-mobile-start/ba-p/14312110); [LeverX SAP Joule Use Cases](https://leverx.com/newsroom/sap-joule-ai-use-cases)

**Scale**
- SAP serves 400,000+ customers in 180 countries; S/4HANA and Joule serve enterprise-tier clients across manufacturing, retail, finance, and public sector
- NRF 2026: SAP announced Storefront MCP server for SAP Commerce Cloud

**Pricing Model**
- Joule is included in SAP Business AI subscriptions bundled with S/4HANA and other Fusion modules; Joule Studio (custom agent building) pricing is part of BTP subscription; exact per-unit pricing is enterprise-custom and quote-driven

**Key Differentiator**
- Deepest integration with SAP's ERP data and business process graph; SAP Knowledge Graph grounding; ABAP code modernization agent uniquely unlocks SAP-specific technical debt reduction; model-agnostic architecture (BYO LLM via SAP AI Core)

**Strategic Direction**
- Transition from Joule as a copilot to full agentic orchestrator with Joule Studio GA; multi-agent A2A interoperability and MCP protocol adoption; NVIDIA GTC 2026 showcase of agentic supply chain planning with cuOpt GPU optimization; Foxconn partnership for manufacturing AI; clean-core + AI-native transformation as strategic positioning
- Source: [NVIDIA GTC Session – SAP](https://www.nvidia.com/gtc/session-catalog/sessions/gtc26-s82080/)

---

### 7. Workday AI

**Current AI Agent Capabilities**
- **Sana AI Agents**: branded suite of purpose-built agents for HR and finance workflows, including:
  - IT Admin Agent, BP Optimize Agent, Deployment Agent
  - Skills identification agents, workforce planning agents
- **Workday Build**: low-code tool for building custom AI agents and agent interactions within Workday platform
- **Agent System of Record**: analytics dashboard for tracking ROI per agent; manages human + digital labor as one workforce system
- 1.7 billion AI actions delivered across the platform in fiscal 2026
- Sana (acquired ~2025): AI-driven learning experience and knowledge; Sana Core and Sana Enterprise in general availability; integrated with Workday HCM for 11,500+ global customers
- Paradox conversational ATS: available through Workday platform for recruiting automation
- Pipedream (acquired ~2025): integration platform supporting AI agent connections across business systems
- **Flowise acquisition** (completed August 14, 2025): low-code agent builder with 42,000+ GitHub stars; visual builder for complex agentic workflows; supports debugging, evaluation, analytics; open-source foundation preserved; now integrated into Workday's customer-facing platform
- Source: [Workday Acquires Flowise Press Release](https://newsroom.workday.com/2025-08-14-Workday-Acquires-Flowise,-Bringing-Powerful-AI-Agent-Builder-Capabilities-to-the-Workday-Platform); [Workday AI Agents Page](https://www.workday.com/en-us/artificial-intelligence/ai-agents.html); [Learning News – Workday FY2026 Results](https://learningnews.com/news/learning-news/2026/workday-deepens-ai-and-learning-push-as-revenues-rise-13)

**Revenue / Scale**
- **11,000+ organizations** worldwide (including 60%+ of Fortune 500)
- FY2026 revenue: **$9.6 billion** (+13.1% YoY); subscription revenues: **$8.8 billion** (+14.5% YoY)
- Q4 subscription revenues: $2.4B (+15.7% YoY); operating cash flows $2.9B (+19.4%)
- Source: [Learning News – Workday FY26 Results](https://learningnews.com/news/learning-news/2026/workday-deepens-ai-and-learning-push-as-revenues-rise-13)

**Pricing Model**
- Subscription-based, enterprise contract pricing (per user / per module, annual terms); quote-driven for all tiers; Sana AI agents included or offered as add-on depending on tier and deployment
- Source: [Workday AI Agents FAQ](https://www.workday.com/en-us/artificial-intelligence/ai-agents.html)

**Key Differentiator**
- Unified HR + Finance platform with AI at core; Agent System of Record for enterprise-wide AI governance; Flowise integration brings developer-grade agent builder to Workday's 11,500+ customer base; real-time skills intelligence and workforce planning AI

**Strategic Direction**
- "AI platform for managing people, money, and agents" (official positioning); aggressive agentic AI roadmap in FY2027+; Flowise powers customer self-service agent building; Sana integration creates AI-native learning within workflow; transitioning from AI features to agentic workforce management
- Source: [Workday Agentic AI Revolution, Feb 2026](https://www.workday.com/en-us/perspectives/ai/2026/02/agentic-ai-revolution.html)

---

## Dedicated Agent / Automation Platforms

### 8. Kore.ai

**Current AI Agent Capabilities**
- **300+ pre-built AI agents** in the Agent Marketplace; enables 10x faster deployment from proven building blocks
- **Multi-agent orchestration engine**: AI specialist agents collaborate with short-term and long-term memory, intelligent delegation, and tool access (Everest Group citation)
- 250+ plug-and-play enterprise connectors (CRM, ITSM, HRIS, ERP, data lakes)
- **Agent Management Platform (AMP)** launched March 17, 2026: unified governance layer for heterogeneous AI environments (LangGraph, CrewAI, AutoGen, Google ADK, AWS AgentCore, Microsoft Foundry, Salesforce Agentforce, proprietary systems)
- No-code + pro-code development: visual drag-and-drop workflow builder alongside full API/SDK extensibility
- Native DialogGPT model; multi-LLM support (XO GPT, third-party, open-source) with lifecycle management and guardrails
- AI governance-first: observability, anomaly detection, cost tracking, policy enforcement across heterogeneous agent environments
- Customers: 400+ Fortune 2000 enterprises including Morgan Stanley, Pfizer, Coca-Cola
- Source: [Kore.ai Agent Management Platform Launch, Mar 17 2026](https://www.kore.ai/news/kore-ai-launches-agent-management-platform); [Kore.ai Best AI Agents for Customer Service](https://www.kore.ai/blog/top-ai-agents-for-customer-service-tested-reviewed)

**Analyst Recognition**
- Leader in Gartner Magic Quadrant for Conversational AI Platforms 2025 (3rd consecutive year); highest Ability to Execute rating
- Leader in Forrester Wave: Cognitive Search Platforms, Q4 2025
- Leader in Everest Group Agentic AI Products PEAK Matrix Assessment 2026
- Leader in AIM Research PeMa Quadrant for Agentic AI Platform Providers 2026
- Source: [Kore.ai Gartner MQ Announcement](https://www.kore.ai/news/kore-ai-named-a-leader-in-gartner-magic-quadrant-for-conversational-ai-platforms)

**Pricing Model**
- Session/activity-based billing for Automation AI and Search AI; per-seat billing for Contact Center AI and Agent AI
- Essential: ~$50/month (annual) — up to 500 conversations/month, basic chatbot
- Advanced: ~$150/month (annual) — up to 5,000–10,000 conversations/month, A/B testing, RBAC
- Standard (pay-as-you-go): $0.20/conversation, $100 minimum
- Enterprise: custom pricing starting ~$300,000/year; dedicated support, on-premise option, unlimited sessions, SSO
- Agent AI: $50–$150/seat/month (named or concurrent)
- Large enterprise: $400,000–$500,000+/year for global deployments
- Source: [Vida.io Kore.ai Pricing Guide](https://vida.io/blog/kore-ai-pricing); [eesel.ai Kore.ai Pricing](https://www.eesel.ai/blog/kore-ai-pricing)

**Key Differentiator**
- Vendor-agnostic Agent Management Platform (AMP) that governs agents across all major frameworks and clouds; broadest coverage of Fortune 2000 enterprises; combines CX, EX, and business process automation in one platform

**Strategic Direction**
- AMP positions Kore.ai as the enterprise control plane for all AI agents regardless of origin; expanding beyond conversational AI into full agentic orchestration and governance; competing directly with enterprise IT/ops platforms for the "agent management" layer
- Source: [Kore.ai AMP Announcement](https://www.kore.ai/news/kore-ai-launches-agent-management-platform)

---

### 9. UiPath

**Current AI Agent Capabilities**
- **Agentic Platform (2026)**: three pillars:
  1. **Autopilot**: cross-platform AI assistant for attended automation
  2. **Maestro**: orchestration layer coordinating humans, bots, and AI agents in complex multi-step processes (BPMN-native)
  3. **ScreenPlay**: computer-use and screen navigation capabilities
- **Agent Builder**: build custom AI agents integrated with existing RPA bots
- **AI Units**: flexible consumption currency for specialized agents, document understanding, computer vision
- UiPath Fusion 2026 user conference (March 2026) showcasing scale deployments
- Healthcare agentic solutions (ViVE 2026, Feb 2026): medical records summarization, claim denial resolution, prior authorization automation
- Peak AI acquisition (early 2026): adds vertical AI agents for retail inventory management and other industry-specific use cases
- WorkFusion acquisition (early 2026): Anti-Money Laundering (AML) and financial crime agents
- GAAP profitability achieved in FY2026 (first full year); ARR approaching $2B milestone
- Management note: agentic capabilities not expected to materially impact FY2026 revenues; investment phase for FY2027 adoption
- Source: [UiPath Healthcare Agentic Solutions, Feb 2026](https://www.uipath.com/newsroom/uipath-launches-agentic-solutions-for-healthcare-vive2026); [Yahoo Finance – PATH Agentic AI, Mar 2026](https://finance.yahoo.com/news/path-agentic-ai-uipaths-platform-164200194.html); [FinancialContent – UiPath Pivot](https://www.financialcontent.com/article/finterra-2026-3-13-the-agentic-era-a-deep-dive-into-uipaths-path-pivot-to-profitability-and-autonomous-systems)

**Pricing Model**
- Unified pricing (credit-based):
  - Basic: **$25/month** (up to 5 users, 2 robots, limited agent calls)
  - Standard: Contact sales (no limit on users/robots, standard features)
  - Enterprise: Contact sales (unlimited, full governance, audit, process mining, AI Center)
- Flex pricing (license-based):
  - Pro: ~$420/month (up to 5 unattended robots)
  - Enterprise: Custom annual
- Consumption units: Platform Units (Unified), or separate AI Units, Robot Units, Agent Units (Flex)
- Source: [UiPath Pricing Page](https://www.uipath.com/pricing); [AIMultiple UiPath Pricing](https://aimultiple.com/uipath-pricing)

**Key Differentiator**
- Largest RPA installed base globally with deepest process automation heritage; Maestro enables true human-bot-agent coordination; governance and compliance tooling (audit logs, SSO, RBAC) enterprise-grade; AI Units consumption model decouples revenue from headcount risk

**Strategic Direction**
- RPA-to-agentic transition: positioning as the enterprise orchestration layer for hybrid human+bot+AI workforces; Maestro as the BPMN-native agentic process engine; vertical agent acquisitions (Peak AI, WorkFusion) for industry-specific deployment; Agentic AI Foundation (AAIF) Gold Member (Feb 24, 2026)
- Source: [UiPath Agentic Automation Trends 2026 Report](https://www.uipath.com/resources/automation-whitepapers/automation-trends-report)

---

### 10. Zapier AI / Make

#### Zapier AI

**Current AI Agent Capabilities**
- **Zapier Agents** (separate product, formerly beta): autonomous agents that research, decide, and act across 8,000+ connected apps
- Zapier Copilot: AI-assisted Zap creation from natural language
- AI-specific integrations: 30+ new AI-specific integrations announced at ZapConnect 2025 (LLM platforms, AI tools, advanced analytics)
- Zaps (deterministic): multi-step, branching logic with 8,000+ integrations
- Agents (agentic): web browsing, data lookup, autonomous behavior triggers; each action = 1 activity

**Pricing (as of March 2026)**
- Zaps (automation):
  - Free: 100 tasks/month
  - Professional: $29.99/month (750 tasks)
  - Team: $103.50/month (2,000 tasks, 25 users, SSO)
  - Enterprise: Custom pricing
- Zapier Agents (separate):
  - Free: 400 activities/month
  - Pro: $50/month (1,500 activities)
- Combined (Zaps + Agents): $80+/month minimum for meaningful use; can reach $150–200/month with Tables, Interfaces, Chatbots add-ons
- Source: [Zapier Pricing Page](https://zapier.com/pricing); [Arahi AI Comparison](https://arahi.ai/blog/arahi-ai-vs-zapier-agents-affordable-ai-automation-for-business-workflows-2025)

**Key Differentiator**: 8,000+ integration breadth; lowest barrier to entry for non-technical automators; brand recognition

#### Make (formerly Integromat)

**Current AI Agent Capabilities**
- **Next-Generation AI Agents** announced Feb 11, 2026: agents embedded directly in the visual automation canvas with full transparency
- Reasoning Panel: real-time view of agent decision-making, tool calls, and branching logic
- Multi-modal support: agents accept and produce PDFs, images, CSVs natively
- Chat directly with agents inside the canvas for real-time testing/refinement
- Pre-built AI agent templates for inventory management, research, triage, reporting
- 3,000+ app integrations within the visual platform
- Full deterministic + agentic combinations shareable across teams as scenario solutions

**Pricing**: Tiered by operations/month; available at make.com; team and enterprise plans with SSO, custom branding

**Key Differentiator**: Transparent agentic reasoning within visual workflow canvas; no "black box" AI decisions — every agent decision visible and auditable

**Strategic Direction (Both)**: Automation-to-agent transition; bridging no-code workflow automation with autonomous AI decision-making; Zapier competing via integrations breadth, Make competing via transparency and visual design
- Source: [Make Next-Gen AI Agents Announcement, Feb 2026](https://www.make.com/en/blog/announcing-next-generation-make-ai-agents)

---

### 11. n8n

**Current AI Agent Capabilities**
- Native **AI Agent node**: autonomous reasoning, tool invocation, iterative planning with memory across conversations; most powerful AI primitive in the platform
- **AI Workflow Builder**: natural language to workflow (text-to-workflow generation)
- 400+ modular, open-source nodes including native connectors for OpenAI, Hugging Face, Cohere, Claude, Gemini, and local LLM providers
- **Self-hosted AI Starter Kit**: open-source template for local AI + self-hosted agent deployment
- Multi-agent workflows: RAG pipelines, classification agents, CRM and Slack integrations, email automation agents
- Supports JavaScript and Python natively in Function nodes alongside visual workflow building
- Queue mode with Redis for horizontal scaling of high-throughput agent workloads
- Community: 4,000+ starter templates; open-source nature enables fastest feature addition in automation tools
- Reddit community assessment (March 2026): "n8n's AI capabilities are at least 6 months ahead of Zapier and Make"
- Source: [n8n.io – Product Page](https://n8n.io); [n8n GitHub Self-Hosted AI Starter Kit](https://github.com/n8n-io/self-hosted-ai-starter-kit); [Reddit Automation Landscape March 2026](https://www.reddit.com/r/SaaS/comments/1ru9tv6/the_automation_tool_landscape_has_changed/)

**Pricing Model**
- **Self-hosted Community Edition**: Free, unlimited workflows and executions, all core features, community support
- Cloud plans:
  - Starter: **€20/month** (billed annually) — 2,500 executions/month
  - Pro: ~$60/month — 10,000 executions, RBAC, insights dashboard
  - Enterprise: Custom — unlimited executions, audit logs, SSO, dedicated SLA
- Note: n8n charges per complete workflow execution (not per step), unlike Zapier
- Self-hosted Business plan (with SSO, Git integration): execution-based licensing fee added in 2025/2026 update
- Typical self-hosted cost: $5–$20/month for infrastructure
- Source: [n8n Pricing Page](https://n8n.io/pricing/); [Zeabur n8n Pricing Guide 2026](https://zeabur.com/blogs/n8n-pricing-shift-self-hosting-business-costs-zeabur-guide); [Lindy n8n Pricing 2026](https://www.lindy.ai/blog/n8n-pricing)

**Key Differentiator**
- Source-available (fair-code license) enables full self-hosting with zero vendor dependency; strongest developer flexibility (code + visual); fastest AI node iteration due to open-source community contributions; data sovereignty for regulated industries

**Strategic Direction**
- Positioning as the developer-first, enterprise-grade automation platform competing on flexibility vs. Zapier/Make's ease-of-use; AI-native workflow builder as default starting point; self-hosted AI Starter Kit enabling private, air-gapped deployments; expected: AI builder availability for self-hosted (decision expected Q1 2026)

---

## Infrastructure / Model Providers

### 12. OpenAI API

**Current AI Agent Capabilities**
- **Agents SDK** (launched March 2025, actively developed): orchestrates multi-agent workflows; supports agent handoffs, input/output guardrails, tracing/observability; compatible with any Chat Completions-style API
- **Responses API** (GA March 2025; Assistants API deprecated Aug 2026): combines chat completions with persistent context, built-in tools (web search, file search, computer use, code interpreter); enables multi-turn, multi-step agentic tasks in a single call
- Web search tool: 90% accuracy on SimpleQA benchmark (GPT-4o search preview)
- Computer use tool: 38.1% on OSWorld (early stage)
- Assistants API officially deprecated August 26, 2026; all developers migrating to Responses API
- GPT-5.4 family: flagship for coding and agentic tasks; GPT-5.4-pro for highest capability
- o4-mini, o3: specialized reasoning for complex STEM/multi-step problems
- ChatGPT Enterprise: HIPAA-compliant, P3/P4 data protection, private model deployments (used by UCSF, major enterprises)
- CFO Sarah Friar (early 2026): pivot toward outcome-based revenue sharing with enterprise customers; moving from token-to-token API pricing toward value-based pricing
- Source: [OpenAI Agents SDK / Responses API – InfoQ, March 2025](https://www.infoq.com/news/2025/03/openai-responses-api-agents-sdk/); [OpenAI Deprecations – Assistants API](https://developers.openai.com/api/docs/deprecations/); [VentureBeat – Responses API Upgrade, Feb 2026](https://venturebeat.com/orchestration/openai-upgrades-its-responses-api-to-support-agent-skills-and-a-complete)

**Pricing (as of March 2026)**
| Model | Input ($/1M) | Output ($/1M) |
|---|---|---|
| gpt-5.4 | $2.50 (short ctx) | $15.00 (short ctx) |
| gpt-5.4-pro | $30.00 | $180.00 |
| gpt-5 (flagship) | $1.25 | $10.00 |
| gpt-5-nano | $0.20 | $1.25 |
| gpt-4o | $2.50 | $10.00 |
| gpt-4o-mini | $0.15 | $0.60 |
| o4-mini | $4.00 | $16.00 |

- Web search: $10/1,000 calls; Batch API: 50% discount; Prompt caching: 50–90% savings
- Enterprise: $20/seat/month (API usage billed separately on consumption)
- Source: [OpenAI API Pricing](https://openai.com/api/pricing/); [CostGoat OpenAI Pricing Calculator](https://costgoat.com/pricing/openai-api)

**Enterprise Position**
- OpenAI holds ~34.4% of business AI subscription market share (Ramp data, Feb 2026) — still #1, down from higher peak
- Daily active users up 10x YoY (Nadella/Altman commentary); 500M+ ChatGPT weekly active users (consumer + enterprise)
- Source: [The Register – Anthropic Catching Up, Mar 2026](https://www.theregister.com/2026/03/19/anthropic_claude_market_share/)

**Key Differentiator**
- Broadest model family (GPT-5.4 → GPT-5-nano); best-in-class computer use and multi-modal tools; largest developer ecosystem; Agents SDK as the de facto agent orchestration framework for many builders

**Strategic Direction**
- Pivot from raw API token sales to outcome-based enterprise revenue sharing; Agents SDK as platform play; deprecating legacy APIs (Assistants) in favor of unified Responses API; expanding enterprise distribution with ChatGPT Enterprise; agentic AI factories / Operator capabilities

---

### 13. Anthropic API

**Current AI Agent Capabilities**
- **Claude 4.5 series** (current generation, released November 2025): Haiku 4.5, Sonnet 4.5, Opus 4.5
- Claude Opus 4.6 (premium), Sonnet 4.6 (active flagship) — pricing and capability updates ongoing in 2026
- 200K token context window (standard); 1M token context available on Sonnet 4.5/4.6
- **Model Context Protocol (MCP)**: co-created by Anthropic, donated to Linux Foundation's Agentic AI Foundation (AAIF) in December 2025; now the de facto standard for LLM-to-tool connectivity; 10,000+ active public MCP servers; 97M+ monthly SDK downloads (Python + TypeScript); adopted by ChatGPT, Cursor, Gemini, VS Code Copilot, Microsoft Copilot
- Claude now has 75+ connectors via MCP directory; Tool Search and Programmatic Tool Calling for production-scale MCP deployment
- **Claude Code**: coding and agent product; estimated ~$2.5B run-rate by early 2026
- Enterprise: HIPAA-compliant, data residency options, SOC 2; Deloitte deployed Claude to 470,000 employees
- Source: [Anthropic MCP Donation to Linux Foundation, Dec 2025](https://www.anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation); [Panto AI – Claude Statistics 2026](https://www.getpanto.ai/blog/claude-ai-statistics)

**Revenue / Scale**
- **300,000+ business customers** (Oct 2025)
- Annualized revenue: **~$14B** (Feb 2026 fundraising disclosure); press reports suggest ~$19B ARR by early March 2026
- **$30 billion Series G** raised February 2026 at **~$380B post-money valuation**
- 500+ customers spending >$1M annually; $100K+ run-rate customers grew 7x YoY
- ~80% of revenue from enterprise/API customers; ~29% enterprise AI assistant market share (analyst estimate)
- New enterprise first-time AI purchases: Anthropic chosen ~70% of the time over OpenAI (Ramp data)
- Anthropic's business subscription share: 24.4% (vs. OpenAI at 34.4%) and gaining fast (+4.9% MoM in Feb 2026 vs. OpenAI -1.5%)
- Source: [The Register – Claude Market Share, Mar 2026](https://www.theregister.com/2026/03/19/anthropic_claude_market_share/); [Panto AI – Claude Statistics](https://www.getpanto.ai/blog/claude-ai-statistics); [Anthropic Economic Index January 2026](https://www.anthropic.com/research/anthropic-economic-index-january-2026-report)

**Pricing (as of March 2026)**
| Model | Input ($/1M) | Output ($/1M) |
|---|---|---|
| Claude Haiku 4.5 | $1.00 | $5.00 |
| Claude Sonnet 4.5/4.6 | $3.00 | $15.00 |
| Claude Opus 4.5 | $5.00 | $25.00 |
| Claude Opus 4.6 (fast mode) | $30.00 | $150.00 |

- Batch API: 50% discount; Prompt caching: cache write ~1.25x input price, cache read ~0.10x
- Long context (>200K tokens): 2x input, 1.5x output premium
- US-only inference: 1.1x multiplier
- Enterprise plans: $20/seat/month (Team Standard); $100/seat/month (Team Premium); API billed separately
- Source: [Silicon Data Claude API Pricing 2026](https://www.silicondata.com/use-cases/anthropic-claude-api-pricing-2026/); [MetaCTO Claude API Pricing](https://www.metacto.com/blogs/anthropic-api-pricing-a-full-breakdown-of-costs-and-integration)

**Key Differentiator**
- MCP protocol ownership and ecosystem (10,000+ servers, cross-platform standard); safety-first positioning attracting regulated industries; Claude Code as the leading AI coding agent; fastest-growing enterprise market share among AI providers

**Strategic Direction**
- MCP as open standard for the agentic internet; Claude Code as primary enterprise growth driver; AAIF participation shaping open agentic AI standards; expanding enterprise partnerships (Salesforce, AWS, Deloitte); safety/alignment research as competitive moat

---

### 14. Google Vertex AI

**Current AI Agent Capabilities**
- **Vertex AI Agent Builder**: full-stack platform for building, scaling, and governing AI agents in production; includes Agent Development Kit (ADK), Agent Engine (runtime), Agent Tools, and Agent Garden (tool registry)
- **Agent Development Kit (ADK)**: now supports TypeScript alongside Python; full Gemini 3 Pro and Flash support; state management with failure recovery, human-in-the-loop pausing, context rewinding
- **Agent Engine** (GA): fully-managed runtime for session and memory management in production; expanded to 7+ additional regions; billing began for sessions and memory bank (Jan 28, 2026)
- **Enhanced Tool Governance**: Cloud API Registry integration in Agent Builder Console; admins govern available tools across the organization; pre-built Google service MCP tools (BigQuery, Google Maps); custom MCP server support via Apigee
- **A2A Protocol (Agent-to-Agent)**: Google-led cross-agent communication standard; supported in ADK + Agent Engine
- **Model Garden**: 200+ models including Gemini, Gemma, Imagen, Veo, Meta Llama, Mistral, Anthropic Claude; 1-click deployment to production endpoints
- Gemini 3 Pro: most powerful model for agentic workflows, multimodal understanding (1M token context, reasoning, coding); available on Model Garden and all Vertex products
- Vertex AI Search, RAG Engine, Code Execution as built-in agent tools
- Security Command Center integration: Threat Detection for agent deployments
- Source: [Google Cloud Agent Builder Documentation](https://docs.cloud.google.com/agent-builder/overview); [Enhanced Tool Governance Blog, Dec 2025](https://cloud.google.com/blog/products/ai-machine-learning/new-enhanced-tool-governance-in-vertex-ai-agent-builder); [Vertex AI Model Garden](https://cloud.google.com/model-garden)

**Pricing**
- Vertex AI Generative AI (token-based):
  - Gemini 3 Pro: $2.00/1M input tokens (≤200K), $12.00/1M output tokens; $4.00/$18.00 for long context
  - Gemini 2.5 Pro: $1.25/1M input, $10.00/1M output
  - Gemini 2.5 Flash: $0.10/1M input, $0.40/1M output
  - Batch API: 50% discount; context caching available
- Agent Engine runtime: compute-based (vCPU + RAM); $0.0864/hour vCPU; $0.009/hour GiB RAM above free tier
- Custom training: from $21.25/hour per node
- Source: [Vertex AI Generative AI Pricing](https://cloud.google.com/vertex-ai/generative-ai/pricing); [Finout Vertex AI Pricing 2026](https://www.finout.io/blog/top-16-vertex-services-in-2026)

**Key Differentiator**
- Google-native model ecosystem (Gemini family) with deepest Google Search grounding; A2A protocol leadership; 200+ model garden for flexibility; enterprise-grade security and governance built into agent infrastructure; strong multimodal capabilities (text, image, video, audio)

**Strategic Direction**
- Vertex AI as the Google Cloud enterprise AI development platform; ADK + Agent Engine as alternatives to Microsoft Foundry and AWS AgentCore; heavy investment in A2A and MCP protocol standardization; Gemini 3 Pro positioning as leading agentic model; 2026 AI Agent Trends report released to guide enterprise adoption

---

### 15. AWS Bedrock

**Current AI Agent Capabilities**
- **Amazon Bedrock AgentCore**: agentic platform for building, deploying, and operating agents at scale with no infrastructure management; components:
  - **Memory**: intelligent long-term and session memory for agents
  - **Gateway**: secure, controlled access to tools and data
  - **Policy** (GA March 2026): centralized, fine-grained governance controls
  - **Runtime**: AG-UI protocol support (GA March 13, 2026); enterprise-grade security and dynamic scaling
- Multi-model access: Anthropic Claude, Meta Llama, Amazon Nova, Mistral, Cohere, and more across 30+ foundation models
- Amazon Nova family: Nova Premier ($0.0025/1K input tokens), Nova Pro ($0.0008), Nova Lite ($0.00006), Nova Micro ($0.000035) — widest cost range; Nova Micro is cheapest on platform
- Knowledge Bases, Guardrails, model evaluation as managed services
- Cross-region inference routing (no additional cost)
- Batch inference: 50% lower cost for asynchronous workloads
- Prompt caching: up to 90% savings on repeated inputs
- Provisioned throughput for guaranteed capacity (reserved hourly billing)
- Source: [Amazon Bedrock AgentCore](https://aws.amazon.com/bedrock/agentcore/); [AWS Bedrock Pricing](https://aws.amazon.com/bedrock/pricing/); [Policy in AgentCore GA, Mar 2026](https://aws.amazon.com/about-aws/whats-new/2026/03/policy-amazon-bedrock-agentcore-generally-available/); [AgentCore AG-UI Protocol, Mar 2026](https://aws.amazon.com/about-aws/whats-new/2026/03/amazon-bedrock-agentcore-runtime-ag-ui-protocol/)

**Pricing (representative on-demand)**
| Model | Input ($/1K) | Output ($/1K) |
|---|---|---|
| Amazon Nova Micro | $0.000035 | $0.00014 |
| Amazon Nova Lite | $0.00006 | $0.00024 |
| Amazon Nova Pro | $0.0008 | $0.0032 |
| Amazon Nova Premier | $0.0025 | $0.0125 |
| Mistral Large 3 | $0.0050 | $0.0150 |

- Agents pricing: token-based per model used for orchestration (each agent step generates input + output tokens)
- Knowledge Base: OpenSearch Serverless from ~$345/month storage minimum
- Source: [GoCloud Amazon Bedrock Pricing 2026](https://go-cloud.io/amazon-bedrock-pricing/)

**Key Differentiator**
- Broadest multi-provider model catalog without AWS lock-in; AgentCore abstracts all infrastructure for enterprise-scale agent deployment; AWS security posture and compliance (14 regions, FedRAMP, HIPAA, SOC 2); Amazon Nova models offer lowest-cost entry points in enterprise cloud

**Strategic Direction**
- AgentCore as AWS's answer to Azure AI Foundry; completing AG-UI protocol for UI-interactive agents; Policy GA for enterprise governance compliance; 1 million NVIDIA GPU deployment across AWS global regions (GTC 2026); positioning Bedrock as the neutral multi-model inference platform for regulated industries

---

### 16. Azure OpenAI Service

**Current AI Agent Capabilities**
- Access to OpenAI's full model family (GPT-5.4, GPT-5, GPT-4o, GPT-4.1, o4-mini, o3) via Microsoft Azure's enterprise-grade cloud infrastructure
- **Microsoft Foundry** (previously Azure AI Foundry): unified agentic and physical AI platform; 11,000+ model ecosystem including OpenAI, Anthropic Claude, Cohere, Mistral, NVIDIA Nemotron
- **Foundry Agent Service** (GA at GTC 2026): build, deploy, and operate AI agents at scale; enhanced observability and voice capabilities; moves from experimentation to production
- NVIDIA Nemotron open models available on Microsoft Foundry for specialized enterprise agent building (GTC 2026)
- FedRAMP and HIPAA compliant; private network access; 99.9% uptime SLA; Azure AD authentication; role-based access control; audit logging
- Provisioned throughput: guaranteed capacity; deployment-based version control
- Content filtering for responsible AI built in
- Deepseek and other open models accessible via Foundry
- Source: [Kellton Azure OpenAI Enterprise Guide](https://www.kellton.com/kellton-tech-blog/azure-openai-enterprise-business-intelligence-automation); [Smartbridge Azure OpenAI Guide, Mar 2026](https://smartbridge.com/azure-openai-service-enterprise-guide/); [Azure AI Foundry Models](https://azure.microsoft.com/en-us/products/ai-foundry/models/openai); [NVIDIA Blog – GTC Microsoft Foundry](https://blogs.nvidia.com/blog/gtc-2026-news/)

**Pricing**
- Azure OpenAI mirrors OpenAI API pricing with Azure infrastructure overhead; provisioned throughput (PTUs):
  - 1-week commit: $1,200/week per GSU
  - 1-month commit: $2,700/month per GSU
  - 3-month commit: $2,400/month per GSU
  - 1-year commit: $2,000/month per GSU
- Token-based consumption pricing also available (same rates as OpenAI API)
- Source: [Vertex AI Pricing comparison context](https://cloud.google.com/vertex-ai/generative-ai/pricing); Azure portal pricing

**Key Differentiator**
- Only path to OpenAI models with full enterprise compliance (FedRAMP, HIPAA, data residency); deepest Microsoft 365 / Azure Active Directory integration; Foundry as multi-vendor model catalog under enterprise governance; NVIDIA partnership for physical AI workloads

**Strategic Direction**
- Microsoft Foundry as the enterprise AI development control plane (competing with Google Vertex AI and AWS Bedrock); multi-model strategy via Foundry reducing dependency on any single model provider; physical AI and robotics blueprints available on Foundry (NVIDIA Cosmos, Alpamayo models); Azure as the preferred cloud for sovereign AI workloads via Azure Government

---

### 17. OpenRouter

**Current AI Agent Capabilities**
- Multi-model inference routing API: single OpenAI-compatible API endpoint routing to **311+ models** (as of March 2026) from Anthropic, OpenAI, Google, DeepSeek, Meta, xAI, MoonshotAI, Mistral, and more
- Automatic fallback and routing: billed only for successful model runs
- Free models available with rate limits (DeepSeek R1, Llama 3.3 70B, Gemma 3, etc.)
- Credit-based system: add credits once, use across all models; no monthly fees, no expiration
- Volume discounts, annual commits, invoicing/POs available for enterprise
- Streaming responses billed same as standard (per token)

**100 Trillion Token Study (Key Findings)**
- In collaboration with a16z, OpenRouter analyzed 100+ trillion tokens of real-world LLM interactions (Nov 2024–Nov 2025)
- Key findings:
  1. **Open-weight models surged**: Chinese open-source models (DeepSeek, Qwen, etc.) reached ~30% of weekly token volume in some weeks; averaged ~13% of weekly volume over the year, from a negligible 1.2% base
  2. **Creative roleplay dominates**: far exceeds the productivity/coding tasks many assumed would dominate
  3. **Agentic inference rising**: multi-step, tool-calling inference is a growing and distinct usage pattern
  4. **"Glass Slipper" effect**: early user cohorts have disproportionately higher long-term retention than later cohorts
  5. **Reasoning models changed inference patterns**: o1's release (Dec 2024) shifted the field from single-pass to multi-step deliberation
- Published as academic paper: arXiv:2601.10088
- Source: [OpenRouter State of AI Report](https://openrouter.ai/state-of-ai); [arXiv 2601.10088](https://arxiv.org/abs/2601.10088); [OpenRouter Announcement Blog](https://openrouter.ai/announcements/the-2025-state-of-ai-report)

**Pricing**
- Pass-through pricing at or near direct API cost (no markup on most major models):
  - Claude Opus 4.6: $5.00/$25.00 per 1M tokens (same as Anthropic direct)
  - Claude Sonnet 4.5: $3.00/$15.00 (same)
  - GPT-5: $1.25/$10.00 per 1M
  - Gemini 3 Flash: $0.50/$3.00 per 1M
  - DeepSeek Chat: $0.32/$0.89 per 1M
  - Grok 4: $3.00/$15.00 per 1M
  - Llama 3.3 70B: $0.10/$0.32 per 1M
- Source: [CostGoat OpenRouter Pricing Calculator, Mar 2026](https://costgoat.com/pricing/openrouter); [OpenRouter Models Page](https://openrouter.ai/models)

**Key Differentiator**
- Widest model access through a single API key and billing relationship; enables model arbitrage (route to cheapest/fastest model per task); critical research dataset (100T tokens) providing unique market intelligence; no lock-in to any provider

**Strategic Direction**
- Positioning as neutral model infrastructure layer; enterprise features (volume discounts, invoicing, fallback routing) for production use; expanding model catalog; academic research partnerships (a16z) to provide market transparency

---

### 18. Together AI / Groq / Fireworks

#### Together AI

**Current Capabilities**
- Serverless inference for 50+ models including Llama 4, DeepSeek-V3.1, Qwen3, Kimi K2.5, Mistral, and gpt-oss variants
- **ATLAS inference system**: runtime-learning techniques for up to 4x LLM inference acceleration with lossless quality
- Dedicated endpoints: customer-specific GPU reservations for consistent throughput
- Fine-tuning: supervised fine-tuning and DPO across model sizes; serving fine-tuned models at base model prices
- Batch API: 50% lower cost for asynchronous large-scale workloads
- GPU cluster access: H100, A100, H200 configurations

**Pricing (serverless, per 1M tokens)**
- Llama 4 Maverick: $0.27 input / $0.85 output
- DeepSeek-V3.1: $0.60 / $1.70
- Qwen3 235B A22B Thinking: $0.65 / $3.00
- gpt-oss-120B: $0.15 / $0.60
- Gemma 3n 4B: $0.02 / $0.04
- Fine-tuning: $0.48–$2.90/hour depending on model size
- Source: [Together AI Pricing](https://www.together.ai/pricing)

**Key Differentiator**: ATLAS system for performance efficiency; open-source model optimization and fine-tuning; industry-leading price-performance for open-weight models; enterprise privacy via dedicated endpoints

---

#### Groq

**Current Capabilities**
- **LPU (Language Processing Unit)**: purpose-built silicon for ultra-fast LLM inference
- Key speeds: GPT-OSS 20B at 1,000 tokens/second; Llama 4 Scout at 594 TPS; Qwen3 32B at 662 TPS; GPT-OSS 120B at 500 TPS
- Batch API: 50% lower cost, 24-hour to 7-day processing window; no impact to standard rate limits
- Built-in tools for GPT-OSS models: Browser Search ($5/1,000 requests), Code Execution Python ($0.18/hour)
- Enterprise and on-premise deployment available via enterprise inquiry
- Models: Llama 4 Scout, GPT-OSS 20B/120B, Qwen3 32B, Kimi K2, Llama 3.3 70B, Whisper speech recognition

**Pricing (per 1M tokens, select models)**
- GPT-OSS 20B 128k: $0.075 input / $0.30 output
- GPT-OSS 120B 128k: $0.15 / $0.60
- Llama 4 Scout: $0.11 / $0.34
- Kimi K2-0905: $1.00 / $3.00
- Llama 3.3 70B: $0.59 / $0.79
- Whisper Large v3: $0.111/hour transcribed; Whisper Large v3 Turbo: $0.04/hour
- Linear, predictable pricing with no elastic pricing spikes
- Source: [Groq Pricing Page](https://groq.com/pricing)

**Key Differentiator**: Fastest raw inference speed among commercial providers (1,000 TPS on 20B models); LPU architecture vs. GPU-based competitors; predictable linear pricing with no hidden costs; ideal for latency-sensitive agentic applications

---

#### Fireworks AI

**Current Capabilities**
- Serverless inference with high rate limits and postpaid billing; $1 free credits on signup
- Dedicated GPU deployment: A100 ($2.90/hr), H100 ($4.00/hr), H200 ($6.00/hr), B200 ($9.00/hr)
- Fine-tuning with LoRA; fine-tuned models served at same cost as base models (no inference cost premium)
- Batch inference: 40% discount vs. serverless
- Image generation (FLUX family, Imagen, Seedream)
- Enterprise: custom SLAs, compliance, dedicated support

**Pricing (per 1M tokens, serverless)**
- Models <4B parameters: $0.10
- 4B–16B: $0.20
- >16B: $0.90
- MoE 0–56B (e.g., Mixtral): $0.50
- DeepSeek V3 family: $0.56 input / $1.68 output
- GLM-5: $1.00 input / $3.20 output
- gpt-oss-120b: $0.15 / $0.60
- Kimi K2.5: $0.60 input / $3.00 output
- Embeddings: $0.008–$0.016/1M (small models)
- Source: [Fireworks AI Pricing](https://fireworks.ai/pricing)

**Key Differentiator**: Fine-tuning without serving cost premium (unique in market); parameter-size-based pricing provides clear cost model; enterprise customization with dedicated GPU reservations; ranked among top 5 inference providers for 2026 (SiliconFlow, Hugging Face, Fireworks, Groq, Cerebras ranking)

**Strategic Direction (All Three)**
- Open-model inference providers competing on price-performance against closed API providers; ATLAS (Together), LPU silicon (Groq), and fine-tuning economics (Fireworks) as distinct moats; serving the agentic AI surge where token volumes are 10–100x higher than conversational use cases; target customers: AI-native companies, developer teams, enterprises with model flexibility requirements
- Source: [SiliconFlow Best Inference Providers 2026](https://www.siliconflow.com/articles/en/the-best-inference-provider-for-llms); [AI Superior Inference Cost Comparison, Mar 2026](https://aisuperior.com/fastest-llm-inference-api-cost/)

---

## Summary Comparison Table

| Platform | Agent Count / Scale | Pricing Entry Point | Key Metric | Strategic Bet |
|---|---|---|---|---|
| Microsoft 365 Copilot | 80% Fortune 500 active agents | $30/user/month | 15M paid seats; 3.3% conversion | Agent Mode in Office apps; multi-agent orchestration |
| Google Workspace AI | 200K+ token context; Flow builder | $11.20/user/month (Standard) | AI Expanded Add-on launched Mar 2026 | Gemini 3 Pro agentic model; Flow workflow builder |
| Salesforce Agentforce | 29K+ deals; 2.4B AWUs | $2/conv → $0.10/action → $125/user | $800M ARR (+169% YoY) | Digital labor; Agentforce 360 + Atlas RE |
| ServiceNow AI | 80B workflows/year baseline | $100–$160+/fulfiller/month | $200M+ AI ACV; Moveworks ($2.85B acq.) | Autonomous Workforce; outcome-guaranteed agents |
| Oracle AI | 1,000+ agents in Fusion Cloud | Included in Fusion subscription | Banking suite: hundreds of agents alone | AI-native ERP; NeMo Agent Toolkit optimization |
| SAP Joule | 400+ AI use cases; 80% coverage target | BTP subscription (enterprise custom) | Joule Studio GA Q1 2026 | Agentic ERP coordinator; A2A interoperability |
| Workday AI | 11,500+ customers; 1.7B AI actions | Enterprise subscription | $9.6B FY26 revenue (+13% YoY) | Sana agents + Flowise builder; Agent System of Record |
| Kore.ai | 300+ pre-built agents; 400+ Fortune 2000 clients | $50/month (Essential) | Gartner MQ Leader (3rd year) | AMP: vendor-agnostic agent governance control plane |
| UiPath | Full RPA bot ecosystem | $25/month (Basic) | GAAP profitability FY2026 | Maestro orchestration; Peak AI + WorkFusion acquisitions |
| Zapier AI / Make | 8,000+ (Zapier) / 3,000+ (Make) integrations | $29.99/month (Zapier Pro) | $80+/month for Zaps + Agents | Automation-to-agent bridge; Make canvas transparency |
| n8n | 400+ nodes; 4,000+ templates | Free (self-hosted) | "6 months ahead" on AI capabilities | Self-hosted, developer-first; open-source AI node ecosystem |
| OpenAI API | 311+ models via Responses API | $0.05/1M (GPT-5-nano) | 34.4% enterprise market share | Responses API + Agents SDK; outcome-based pricing pivot |
| Anthropic API | 10,000+ MCP servers ecosystem | $1/1M (Haiku 4.5) | $14B ARR; $380B valuation; 300K+ business customers | MCP standard; Claude Code $2.5B run-rate |
| Google Vertex AI | 200+ models in Model Garden | $0.10/1M (Gemini 2.5 Flash) | Agent Engine GA; 7+ new regions | ADK + Agent Engine; A2A protocol; tool governance |
| AWS Bedrock | 30+ foundation models; AgentCore | $0.000035/1K (Nova Micro) | AgentCore Policy GA March 2026 | AgentCore as managed agent infrastructure layer |
| Azure OpenAI | 11,000+ models via Foundry | Token-based (same as OpenAI) | Foundry Agent Service GA (GTC 2026) | Microsoft Foundry multi-vendor platform; compliance moat |
| OpenRouter | 311+ models, single API | $0 (free models) | 100T token study published Jan 2026 | Model arbitrage layer; neutral market intelligence |
| Together AI / Groq / Fireworks | Open-weight inference specialists | $0.02–$9.00/GPU-hr or per-token | Groq: 1,000 TPS; ATLAS: 4x speedup | Speed (Groq), fine-tuning economics (Fireworks), ATLAS perf (Together) |

---

*Registry compiled March 22, 2026. All data sourced from primary vendor disclosures, official earnings releases, and verified press coverage. Source URLs included inline with each claim. Pricing and metrics subject to change; verify directly with vendors for purchasing decisions.*