Evaluating the impact of AI Agents, one job at a time.

The agentic AI landscape has rapidly evolved into a multi-billion dollar ecosystem, with major platforms launching transformative updates throughout 2025. Unlike traditional AI assistants that merely respond to prompts, agentic AI platforms create autonomous systems capable of multi-step reasoning, decision-making, and task execution across enterprise environments. This blog post covers some of the recent breakthroughs in this space and captures the current state of this fast moving market.

Market Overview

The AI agent market reached $5.4 billion in 2024 and is projected to grow at 45.8% annually through 2030, driven by enterprises seeking to automate complex workflows beyond basic chatbot capabilities. Our global survey on GenAI/ AI Agents adoption showed plans for increased investments in AI Agents in 2025 than 2024. According to the survey, the number of respondents investing in Agentic AI increased across all investment buckets, and more sepcifically, respondents planning on investing more than $1M increased by 31%.

Source: Enterprise GenAI/ AI Agents Adoption Survey, Q1 2025, CloudDon

Recent Breakthroughs

Amazon Bedrock AgentCore

Amazon announced Amazon Bedrock AgentCore at the AWS Summit New York 2025, enabling organizations to deploy and operate secure AI agents at enterprise scale with seven core services, including:

  • AgentCore Runtime: Framework agnostic runtime to handle variable workloads that provides serverless environments with session isolation
  • AgentCore Identity: AIAgent authentication that enables them to secure AWS services and third-party tools with pre-authorized user content
  • AgentCore Identity: Secure agent authentication with existing identity providers including Amazon Cognito, Microsoft Entra ID, and Okta
  • AgentCore Gateway: Secure tool discovery and API transformation capabilities
  • AgentCore Code Interpreter: Secure sandbox execution for complex calculations and data processing
  • AgentCore Observability: Comprehensive monitoring and debugging capabilities
  • AgentCore Browser: Web brower to support web web-based automation tasks such scraping a website, form completion, etc.

AWS aims to enable enterprises overcome most cited challenges with AI Agent adoption — cost, techncal complexity, and data privacy through Amazon Bedrock AgentCore. While the scale and security that AWS is known for are in its favor, the success of this offering largely depends on how easy it is deploy (technical complexity) and how cost-effective it is with production use cases.

OpenAI ChatGPT Agent Mode

OpenAI introduced “ChatGPT Agent” on July 17, 2025, available to Pro, Plus, and Team users (with Enterprise and Education access coming soon). This unified agent leverages a virtual sandboxed computer environment to autonomously manage complex workflows — from secure web browsing and user‑authorized logins, to code execution, research synthesis, and creation of editable documents like slides and spreadsheets.

It includes a comprehensive safety stack: user confirmation prompts for consequential actions, dual-use refusal training, real‑time monitoring, threat modeling, classifiers, disabled long‑term memory, and “Watch Mode.” These systems are designed to prevent unauthorized or high‑risk behavior and mitigate misuse in sensitive areas.

AI Agent Landscape

Google Cloud’s Agentspace

Google Cloud’s Agentspace, launched in early 2025, delivers a horizontal orchestration layer that connects across enterprise systems. It features a no‑code Agent Designer (Agent Assembler) and enables agents to execute complex workflows — planning, researching, generating content, and acting on enterprise data — while maintaining neutrality across systems.

Microsoft’s Agentic AI Portfolio

At Microsoft Build 2025 (May 2025), GitHub Copilot unveiled its agent mode, becoming a task‑executing coding partner. Azure AI Foundry now offers access to 1,900+ AI models from Microsoft and partners and supports multi‑agent orchestration via Copilot Studio and Azure Foundry Agent Service — leveraging open protocols like MCP and A2A in frameworks informed by AutoGen research.

Salesforce Agentforce Evolution

Salesforce unveiled Agentforce 3 on June 23, 2025, a major evolution of its digital labor platform. The release introduces a new Agentforce Command Center — a unified observability hub that lets enterprises monitor and manage AI agent performance in real time across metrics like latency, success rates, hand-off frequency, and error trends. The platform now supports six globally available languages (French, Italian, German, Spanish, Japanese, Portuguese) and is live in India, Brazil, the U.K., Canada, and Japan.

ServiceNow’s AI Platform

ServiceNow introduced its AI Platform at Knowledge 2025 (May 6, 2025) — a reimagined Now Platform designed for enterprise-scale agentic AI. The platform includes thousands of pre-built agents for IT, HR, customer service, and industry workflows, with integration support for partner-built agents. The platform includes AI Control Tower for centralized management, AI Agent Fabric for communication backbone, Agent Orchestrator for coordinating teams of agents, and Agent Studio for low-code agent building using natural language prompts.

UiPath’s Agentic Automation Platform

UiPath launched its Agentic Automation Platform on April 30, 2025, unifying AI agents, robotic process automation (RPA), and human participants within a single orchestration layer. Features include Agent Builder for no‑code agent creation, Maestro for workflow orchestration across robots and humans, and a Healing Agent that adapts automatically to interface changes — enabling resilient, natural‑language driven automation.

IBM Watsonx Orchestrate

IBM’s watsonx Orchestrate provides pre‑built agents for HR, procurement, sales, and customer care; customizable via Agent Builder and Flow Builder; and supports multi‑agent teams through its Orchestrator Agent. The Agent Connect framework enables third‑party agents to integrate deeply across enterprise systems.

Open Source Innovations

  • LangChain remains a top choice for developers building conversational agents and retrieval-augmented systems. Its modular design — with interchangeable chains, agents, tools, and memory modules — allows you to flexibly compose workflows across multiple LLM providers and data sources.
  • CrewAI is a mature open-source multi-agent orchestration framework that empowers teams to build collaborative AI workflows — agents assigned defined roles and goals can interact, delegate, and execute multi-step tasks. It is model-agnostic, deployable anywhere, and backed by a rapidly expanding developer community

Specialized Solutions and Market Positioning

  • Enterprise platforms such as Salesforce Agentforce, ServiceNow AI Platform, UiPath Agentic Platform, and Moveworks lead in product maturity, backed by proven deployments and strong domain-specific performance. However, their flexibility remains constrained by ecosystem lock-ins, prioritizing stability and ROI over open customization.
  • Hyperscaler offerings — including Microsoft’s Agentic AI portfolio, AWS AgentCore, Google Agentspace, and IBM Watsonx Orchestrate — offer broader architectural openness with varying maturity levels:
    —Microsoft stands out for its high maturity and flexibility, combining Copilot’s ubiquity with Azure AI Foundry’s multi-LLM orchestration.
     — Google Agentspace distinguishes itself with hybrid and multicloud support, enabling orchestration across diverse enterprise data sources and systems, while maintaining strong AI and data integration capabilities.
     — AWS AgentCore delivers model-agnostic agent orchestration anchored in Bedrock and AWS’s enterprise footprint.
     — IBM Watsonx Orchestrate focuses on workflow-specific maturity (HR, procurement) with mid-range flexibility through AI Gateway integrations.
  • Open frameworks like LangChain and CrewAI deliver maximum flexibility, supporting multi-LLM, multi-agent orchestration. However, they remain less mature for enterprise production — with LangChain more established and CrewAI still early-stage but rapidly evolving.
  • Finally, specialized tools such as Cognosys (browser-native automation) and Relevance AI (experimental workflow automation) emphasize niche innovation, trading breadth and enterprise readiness for targeted adaptability.

Agentic AI Platforms — Product Maturity vs Flexibility © CloudDon

Recommendations for Buyers

As the agentic AI platform landscape continues to evolve rapidly — spanning enterprise suites, hyperscaler ecosystems, open frameworks, and specialized tools — buyers should approach selection strategically, balancing integration, scalability, and governance considerations.

1. Prioritize Integration and Interoperability

Select platforms that support open standards and protocols (such as MCP and A2A), provide extensive pre-built connectors, and demonstrate seamless integration with your existing enterprise systems. Emphasize solutions that reduce complexity and fit naturally within your current technology landscape. Avoid tools that create rigid vendor lock-in or require significant re-architecture to deploy.

2. Start with Proven Use Cases & Scale Gradually

Begin with pre-built agents tailored to common enterprise functions (e.g., HR, procurement, customer service) or your specific industry rather than attempting to build entirely custom agents from scratch. Favor platforms that enable a controlled expansion of agent autonomy, allowing you to incrementally increase capabilities based on demonstrated performance and trust-building with human oversight.

3. Evaluate Total Cost of Ownership Holistically

Assess the end-to-end cost implications, not just subscription costs. Consider factors such as integration effort, ongoing maintenance, governance overhead, compliance requirements, and the internal expertise needed to manage and scale agents effectively.

  • Enterprise platforms often entail higher upfront costs but include built-in compliance, security, and support that can reduce operational burden.
  • Open frameworks or toolkits provide greater customization potential but require in-house development skills, stronger governance controls, and more intensive operational oversight.

Conclusion

The agentic AI landscape in 2025 reflects intense competition across enterprise-focused platforms, hyperscaler solutions, and open-source alternatives. Recent launches of Amazon Bedrock AgentCore and OpenAI’s ChatGPT Agent mode underscore the accelerating innovation pace. Each platform addresses distinct market needs while advancing autonomous AI system adoption across enterprise environments, with clear differentiation emerging between proven specialists and platforms still establishing market presence.

Three dominant strategies define this landscape:

  • Enterprise workflow providers capitalizing on established customer bases and automation expertise to deliver domain-specific, outcome-driven agents.
  • Hyperscalers enabling horizontal orchestration and hybrid deployment at scale, leveraging deep infrastructure and ecosystem reach.
  • Open-source frameworks prioritizing flexibility and extensibility for organizations with strong technical capabilities and custom integration needs.

Ultimately, success in adopting agentic AI will hinge on an organization’s ability to balance enterprise-grade maturity with the flexibility required for evolving use cases, ensuring both immediate impact and long-term adaptability.