In 2025, the imperative was simply to adopt agentic AI or get left behind. But in 2026, the reality has shifted: agents are exploding across the enterprise from every vector.
From internal developers leveraging new frameworks to legacy applications quietly adding agentic features in patch updates, the volume of agents has grown from dozens to thousands.
In our recent webinar, The Agent Explosion is Here, Arthur CEO Adam Wenchel broke down how to navigate this shift by building a rock-solid Agent Discovery & Governance (ADG) strategy.
"Trying to manually track the thousands of agents entering your enterprise is not a winning game. You need automated discovery to find what's running, followed by a rock-solid governance strategy that scales beyond human oversight to manage the risks." — Adam Wenchel
In this webinar, Adam focused on the a few key principles required to manage this new agent explosion:
- Implement the Agent Development Lifecycle (ADLC): Unlike traditional deterministic software cycles, agent development is more like a continuous flywheel. Success in agentic development requires moving beyond "vibes-based" iteration to a methodical approach that establishes baselines and rigorous evaluation loops before production.
- Identify "Shadow Agents": The explosion of agents is driven by three distinct vectors that every enterprise must monitor:
- Internal Development: Application teams actively building new software using modern agentic frameworks.
- New Vendor Solutions: Purpose-built tools from AI-first startups designed to automate specific verticals like legal, finance, or customer service use agents under the hood.
- Legacy Software Updates: Traditional applications (like CRMs / Ledgers) are weaving agents into their software through routine patches and updates.
- Automate Discovery: You cannot govern what you cannot see. Effective discovery requires a multi-layered approach to catch agents as they come online:
- Telemetry (OTEL): Standardizing agent tracing to use OTEL allows you to discover agents and tools directly from the log streams.
- MCP Monitoring: Tracking Model Context Protocol (MCP) servers helps identify when new agents or tools are exposed to other applications.
- Network Layer Analysis: Analyzing HTTP traffic bodies helps spot new LLM usage patterns and agent communications at the network level.
- API-Driven Discovery: Leveraging built-in hooks from platforms like AWS Bedrock or Google’s Vertex AI provides visibility into what is running in your cloud environments.
- Governance is a core Agent Development Activity: Governance for AI agents is moving from compliance teams directly to the application developers. To implement an effective governance strategy, policies must be platform-agnostic and highly customizable based on the use-case of the agent. (For instance, "toxicity" guardrail and “friendly tone” eval for a customer support agent looks very different from one for a warehouse’s internal inventory management agent.)
To get the full breakdown on discovery techniques and successful governance frameworks, watch the entire webinar on-demand:
If you’d like to learn more about Agent Discovery & Governance, book a demo with us.
SHARE
