The Agentic Shift: Key Takeaways from IDC Directions 2026

Last week I attended IDC Directions 2026 in Boston, and the overarching message from the conference is that we have moved beyond the AI experimentation phase into a high-stakes era of enterprise-wide agentic execution.  The industry is at a tipping point where the focus has shifted from building AI capacity to driving massive enterprise adoption.

Here are the four trends from the conference that will influence the coming years and what those trends mean for observability.

The Rise of the Agent Economy

By 2029, IDC forecasts that enterprises will collectively run more than one billion AI agents. This represents a reset of the enterprise software model. For observability, it means evolution into becoming an essential component for the orchestration layer of autonomous agents.

Traditional monitoring cannot capture why an agent made a specific decision. Agentic observability becomes important for clarity into multi-agent collaborations. Also, observability needs to address not only if the agentic environments are reliable but if they are behaving as expected, and if they are driving expected business outcomes.

The Four Pillars of Agentic Success

I found IDC’s Alessandro Perilli’s session in the keynote very insightful where he highlighted what is needed to be prepared for the Agent Economy. He shared IDC research on four pillars of agentic success and those are: Knowledge Fabric, Agent Orchestration, Decision Intelligence, and Agentic AI Security.

These pillars align with some of the required capabilities of a modern observability platform. To master the Knowledge Fabric, observability must move beyond MELT to provide deep contextual clarity, mapping how knowledge graphs and semantic definitions influence an agent's reasoning. 

As we scale Agent Orchestration, the next generation of distributed tracing must evolve into agentic tracing or agentic observability to visualize the complex hand-offs between multiple autonomous agents and eliminate black box failures. 

Enabling true Decision Intelligence requires a shift from a human-in-the-loop to a human-on-the-loop model, intelligent observability acts as a supervisor to ensure AI stays within its operational guardrails.

Finally, Agentic AI Security is essential for reliable adoption of AI in the enterprise. This means a desirable convergence of observability with governance, demanding active monitoring for hallucinations, prompt injections, and data leaks to ensure that system reliability is synonymous with security.

The Shift to AI-Ready Infrastructure 

One of key trends is that this AI supercycle is projected to exceed more than $1T in infrastructure spending by 2029. This build-out phase requires a complete modernization of datacenters to handle unprecedented power density and cooling requirements for AI factories. This massive scale of AI-related compute and infrastructure investment is unprecedented in enterprise IT, demanding a new era of end-to-end observability to manage the massive complexity and ensure peak performance.

Managing this scale requires a comprehensive intelligent observability across all levels of agentic performance, software, networks and infrastructure at scale. This is only possible with a unified data layer and intelligent observability built on highly scalable architectures.

Trust, Governance, and Risk

 A key trend highlighted at the conference was agents becoming force multipliers for the workforce. There needs to be a change in how we govern, observe, and monetize all digital interactions.  As AI becomes embedded, traditional guardrails are no longer sufficient. CEOs expect AI implementation risks from synthetic identity attacks to prompt injection to have a major impact this year.

Governance must be prioritized and focusing on the highest-exposure critical user journeys first. Observability is non-negotiable and digital experience management tied to business outcomes are essential indicators of potential risk. For proper governance, one must understand agentic systems and models used across an organization to prevent so-called “shadow AI sprawl" as potential vulnerabilities and organizational risk.

Looking Ahead

The keynote concluded with the introduction of IDC Quanta, an AI-powered intelligence layer that embeds research directly into customer workflows via the Model Context Protocol (MCP). The initiative includes a collaboration with Anthropic to allow users to access IDC's intelligence directly within Claude workflows via MCP and plugins. I saw an early demo of IDC Quanta and it was impressive. I am excited to give it a try once the beta queue opens.

At New Relic, we’ve already embraced this future of open agentic communication by having our own MCP Server and many agentic integrations. We allow enterprises to bring their observability data directly into environments for team use with Cursor or GitHub Copilot or ServiceNow to resolve issues. The open nature of New Relic intelligent observability and remediation focus was also highlighted in IDC’s research where New Relic was recently positioned in the Leaders category in the  IDC MarketScape: Worldwide AIOps 2026 Vendor Assessment (Doc #US54116226, March 2026) report.

We are moving from a world where we observe systems to one where we help orchestrate intelligence and drive outcomes.