The Agentic Platform Landscape, May 2026

The agentic AI space has consolidated dramatically over the last 12-18 months. What was a chaotic field of dozens of frameworks has settled into roughly two layers that big enterprises are actually embedding in their code: open-source orchestration frameworks (the runtime "guts" engineers import into their codebase) and hyperscaler platforms (the managed environments where those agents live, get governed, and connect to enterprise data). Most large corporations are adopting one of each.

Here's what the landscape looks like right now.

The orchestration frameworks (what engineers import into their code)

LangGraph (from LangChain)

LangGraph has emerged as the de facto standard for stateful, production-grade agents. It hit v1.0 in October 2025 and is now the default runtime for all LangChain agents. Its graph-based architecture — nodes, edges, and persistent state instead of linear chains — is what handles retries, checkpointing, human-in-the-loop pauses, and recovery from failed API calls.

Around 400 enterprises run LangGraph in production, including Klarna, Uber, LinkedIn, BlackRock, Cisco, Elastic, JPMorgan, and Replit. Concrete deployments people often cite:

Klarna's AI Assistant, powered by LangGraph and LangSmith, handles customer support for 85 million active users — reducing customer resolution time by 80%.
LinkedIn's AI recruiter uses a hierarchical agent system powered by LangGraph.
Uber's Developer Platform team uses it for large-scale code migrations and automated unit test generation.
Elastic orchestrates threat-detection agents on it.

LangGraph is MIT-licensed open source, with a paid LangGraph Platform for managed deployment and LangSmith for observability/tracing. It's the choice when the cost of a wrong agent decision is high — regulated industries, long-running workflows, anything auditable.

CrewAI

CrewAI is the other half of the open-source story, and it's growing the fastest by adoption count. CrewAI raised $18M Series A, hit $3.2M revenue by July 2025, runs 100,000+ agent executions per day, and 60% of Fortune 500 companies now use it.

Its abstraction is "role-based crews" — you literally define agents with a role, goal, and backstory, and they collaborate. Setup takes 2-4 hours versus a 1-2 week ramp for LangGraph, which is why it dominates for proof-of-concept work and content/analysis workflows. IBM Consulting implemented CrewAI to automate federal eligibility determinations, integrating with WatsonX foundation-model runtime.

The pattern many enterprise teams have settled into: prototype in CrewAI, then rewrite the production path in LangGraph once the design is confirmed and you need tighter state management.

Microsoft Agent Framework

This is the third major framework, and it's significant because Microsoft consolidated AutoGen and Semantic Kernel into one SDK that reached v1.0 general availability in April 2026. This is the default for .NET and Azure-native teams.

OpenAI Agents SDK

Got a major overhaul in April 2026 adding native sandboxing, sub-agents, Codex-style filesystem tools, and first-class MCP support — it's the lowest-friction option if you're already all-in on GPT models.

The hyperscaler platforms (where agents are governed and deployed)

This is where the "big huge corporations" angle really shows. The three cloud giants are now in direct head-to-head competition, and they're approaching it differently.

Microsoft Copilot Studio + Azure AI Foundry + Agent 365

The most widely deployed by raw volume: 160,000 organizations running 400,000+ custom agents, covering roughly 1B Microsoft 365 seats. Microsoft's stack splits into:

Copilot Studio — low-code, for business analysts
Azure AI Foundry Agent Service — code-first, for engineers
Agent 365 — the governance/control plane, launching as part of the M365 E7 bundle

They added MCP server tool support in preview during the April 2026 updates, so Copilot Studio agents can now invoke any MCP-compliant tool.

Salesforce Agentforce

Salesforce is closing Agentforce deals at 29,000 since launch with $800M ARR. Agentforce and Data 360 combined ARR hit nearly $1.4 billion by Q3 FY26, up 114% year-over-year.

Its Spring '26 release introduced Agentforce Builder (a conversational workspace for building agents) and Agentic Enterprise Search (pulling context from 200+ external sources). Summer '26 adds multi-agent orchestration. This is the play for any organization already deep in the Salesforce ecosystem — agents inherit the Salesforce permissions model and execute inside your org.

Google Gemini Enterprise Agent Platform

Launched April 22, 2026 at Google Cloud Next. Google made a significant move by absorbing Vertex AI into this platform — rebranding Vertex AI as Gemini Enterprise Platform.

The notable thing is multi-model support: Model Garden gives access to 200+ models including Gemini 3.1 Pro, Gemini 3.1 Flash Image, Lyria 3, open models like Gemma 4, plus Anthropic's Claude Opus, Sonnet, and Haiku in the same platform. L'Oréal is a publicly cited customer using it.

AWS Bedrock AgentCore

AgentCore is model-agnostic, working with foundation models including OpenAI, Google's Gemini, Anthropic's Claude, Amazon Nova, Meta Llama, and Mistral.

Critically for the "embed in your own code base" angle: AgentCore Runtime is a serverless runtime built for deploying agents using any open-source framework — CrewAI, LangGraph, LlamaIndex, Google ADK, OpenAI Agents SDK, or Strands Agents — with any protocol like MCP or A2A.

AWS's strategy is framework-neutral execution infrastructure; they want to be the runtime no matter what framework you wrote your agent in. AWS is optimizing for velocity with harnesses that bring agents to production faster, while Google's approach is governance-focused with a Kubernetes-style control plane.

What's actually different in 2026 versus a year ago

A few structural shifts are worth flagging.

The agent washing problem is now openly acknowledged. Most vendors are rebranding existing chatbots, RPA scripts, and linear workflow tools as agents — a pattern practitioners call agent washing. Genuine agentic AI requires autonomous decision-making, multi-step reasoning, and dynamic error handling; most products on the market today do not clear that bar.

The adoption gap is real. Despite the hype, only 17% of organizations have deployed AI agents to date, yet more than 60% expect to do so within the next two years — the most aggressive adoption curve among all emerging technologies measured. Most deployments remain narrowly scoped, and fully autonomous agents are not ready for the majority of enterprise use cases.

The dominant failure mode: organizations deploying agents across 10 workflows before validating that any single one delivers consistent value.

Interoperability protocols have matured. MCP (Model Context Protocol, originally from Anthropic) and A2A (agent-to-agent) are now first-class citizens across every major platform. This is the biggest reason enterprises feel safer betting on a framework — you're less locked in than you would have been in 2024.

The practical pattern big corporations are following

When you hear about a Fortune 500 company "taking an agentic platform into their code base," what's almost always actually happening is a two-layer choice:

Pick an open-source framework (LangGraph for stateful regulated workflows, CrewAI for role-based collaboration, Microsoft Agent Framework if .NET-shop) and import it into application code.
Pick a managed runtime/governance platform (AgentCore, Gemini Enterprise, Copilot Studio + Agent 365, or Agentforce if Salesforce-heavy) to handle deployment, identity, observability, and connections to enterprise systems.

The frameworks are converging on similar abstractions, and the platforms are converging on framework-neutrality (AgentCore explicitly runs any of them). The real differentiation is now around governance, observability, and data integration — not the agent loop itself.

Quick reference table

Layer	Option	Best for	Notable adopters
Framework	LangGraph	Stateful, regulated, auditable workflows	Klarna, Uber, LinkedIn, JPMorgan, BlackRock, Cisco, Elastic, Replit
Framework	CrewAI	Role-based multi-agent, fast prototyping	60% of Fortune 500, IBM Consulting
Framework	Microsoft Agent Framework	.NET / Azure-native teams	Microsoft ecosystem shops
Framework	OpenAI Agents SDK	GPT-centric, low friction	OpenAI-aligned teams
Platform	Microsoft Copilot Studio + Agent 365	Microsoft 365 / Dynamics shops	160k orgs, 400k+ agents
Platform	Salesforce Agentforce	Salesforce-heavy orgs	29k deals, $800M ARR
Platform	Google Gemini Enterprise	Multi-model, governance-first	L'Oréal
Platform	AWS Bedrock AgentCore	Framework-neutral runtime	Multi-framework deployments