Strong Signals

The September edition of the Agentic AI Radar confirms a broader trend: the best-positioned solutions are now those that can industrialize AI within governed environments, rather than simply achieve rapid adoption.

Organizations are placing greater emphasis on traceability, identity management, compliance, usage controls, and deployment oversight.

This evolution is benefiting large integrated U.S. platforms, while also increasing concerns around dependence on a limited number of proprietary ecosystems and digital sovereignty.

This shift is reflected in the Radar’s most balanced positions: Microsoft Copilot Studio combines a maturity score of 8.67 with a momentum score of 7.85, while IBM watsonx Orchestrate reaches 7.71 and 6.73. Databricks Agent Bricks, AWS Bedrock AgentCore, Salesforce Agentforce, and GitHub Copilot also stand out. Together, these players form a group of mature solutions that continue to maintain strong market momentum.

Conversely, the Radar shows that momentum alone is no longer enough to define a Leader: the strongest positions now combine adoption with the ability to operate effectively within an enterprise environment.

Signals to Watch

AI Gateways are becoming a new battleground

The emergence of AI Gateways is one of the most significant developments this month. Cloud, cybersecurity, data, and API management players are converging around the same ambition: becoming the central control point between enterprise information systems and AI.

This new layer concentrates security, compliance, and cost-control requirements. It is giving rise to a software category of its own, while also creating a risk of fragmentation if multiple control mechanisms coexist within the same organization.

The Radar reflects this competition. Microsoft now brings together several highly mature offerings, including Copilot Studio, Agent Framework, and Foundry Agent Service, while Agent 365 records the highest momentum score in the dataset. IBM watsonx Orchestrate, Databricks Agent Bricks, AWS Bedrock AgentCore, Salesforce Agentforce, and Boomi AgentStudio illustrate the same trend: platforms are seeking to become the integration point between agents, data, applications, and governance policies.

China is accelerating its platform strategy

Chinese solutions continue to gain ground by focusing on open platforms and large-scale model deployment. Their objective remains consistent: capture adoption quickly before progressively strengthening governance mechanisms.

This momentum is not yet reflected in their maturity scores. Transparency, auditability, and compliance remain their main weaknesses. However, the gap between this acceleration and their current level of trustworthiness is a signal worth monitoring.

The Radar results mainly confirm a gap between platform ambition and currently observed maturity levels: Alibaba’s Qwen-Agent has the strongest momentum within this group at 5.70, but remains low in maturity. ByteDance Coze scores 3.20 in momentum and only 1.91 in maturity; Baidu Qianfan scores 1.47 and 1.83; and Alibaba AgentScope scores 0.24 and 0.64.

These positions do not yet represent a breakthrough among the Radar’s leading platforms. They do, however, highlight the catch-up potential of players that are prioritizing adoption and ecosystem expansion first.

Canada and Europe are betting on sovereignty

Faced with U.S. dominance and China’s acceleration, European and Canadian players are pursuing a distinct path centered on digital sovereignty, regulatory compliance, and data control.

This strategy generates more limited traction, but addresses growing organizational concerns around resilience and technological independence.

Within the Radar, this sovereignty-driven approach has not yet translated into a cohesive European or Canadian group comparable to the leading integrated U.S. platforms. Dataiku AI Agents scores 4.59 in maturity and 4.98 in momentum; Canada-based Cohere Command scores 3.13 and 4.53; and Mistral Agents API scores 2.30 and 2.94. Augure stands out as a notable exception, with a momentum score of 6.83.

The signal is therefore one of a credible alternative in contexts where data residency, compliance, reversibility, and independence from major technology ecosystems are key selection criteria.

Interoperability is becoming a prerequisite

Interoperability mechanisms between agents, models, and tools are rapidly becoming widespread. What was recently a competitive advantage is progressively becoming a basic requirement.

This shift is particularly visible among API management and data platform vendors, which are extending existing governance mechanisms to agentic environments. Providers that fail to follow this trend risk becoming less relevant within enterprise architectures.

This evolution can be seen in the range of solutions gaining traction around orchestration and integration with existing environments:

• LangChain records a momentum score of 7.04.
• LangGraph combines a maturity score of 6.39 with a momentum score of 5.79.
• Dify reaches 8.62 in momentum, the second-highest score in the Radar.

Among enterprise platforms, ServiceNow, UiPath, Zapier, and n8n confirm that the focus is shifting toward integrating agents into workflows, business applications, and enterprise information systems.

The market is becoming increasingly polarized

The gap between adoption and maturity continues to widen. Many solutions show strong momentum without yet providing all the safeguards required for large-scale deployment, while more mature platforms are showing signs of slower growth.

This polarization reflects a reality already visible in the market: AI adoption is often moving faster than organizations’ ability to govern it and deploy it effectively in production.

The Radar further illustrates this polarization:

• Some solutions benefit from strong traction while still showing more limited maturity, including Dify, Google Agent Development Kit, Uniphore Business AI Cloud, Devin, Hippocratic AI, and Qodo.
• At the opposite end, several already mature offerings are progressing more slowly, including Mastra, Pydantic AI, and Claude Agent SDK.

This divergence is critical when considering long-term use cases, where technology choices are made with a multi-year horizon in mind.

Early Signals

Local agent execution is becoming viable

Advances in open models and specialized hardware are making local or hybrid agent execution increasingly realistic. Platforms will now need to manage a continuum spanning cloud, on-premises infrastructure, and edge environments, while maintaining security, model distribution, and usage traceability.

Agent identity is becoming a prerequisite

As agents become more autonomous, the ability to assign them a clear identity, mandate, and set of permissions is becoming essential. Access control, delegation, and authentication mechanisms are likely to play an increasingly important role in agentic platforms.

Continuous evaluation is becoming a production capability

Agent quality can no longer be assessed only before deployment. Platforms are evolving toward continuous monitoring mechanisms that make it possible to test, compare, and correct agent behavior during operation, much like the practices already established for traditional software.

The workplace is becoming a new agentic environment

Agents that can interact directly with applications, files, and browsers are opening up new automation opportunities. They also introduce new challenges related to security, privilege management, and oversight of actions performed directly on employee workstations.