From assistants to agents: Lessons on agentic enterprise at Gartner IT Symposium & Xpo

At this year’s Gartner IT Symposium/Xpo, CIOs, IT leaders, and digital workplace innovators gathered to explore the technologies shaping the next era of enterprise productivity. The event’s central theme — the rise of the agentic enterprise — sparked a crucial conversation about where AI truly delivers business value.

Across sessions, Gartner analysts, vendors, and executives agreed: AI’s next chapter isn’t just about conversations, it’s about collaboration.

And the enterprises that evolve from static automation to adaptive, agentic intelligence will define the future of work.

The agentic shift: From assistants to autonomous actors

Gartner’s “Six Categories of AI Agent Capabilities” captured a concept every CIO needs to understand: not all AI agents are created equal.

AI assistants operate at the minimal or emerging levels of maturity—performing defined, reactive tasks through signals or deterministic logic. But agentic AI introduces autonomy, decisioning, and learning. These systems don’t just respond; they act, optimize, and evolve within complex environments.

The framework breaks AI capabilities into six key dimensions:

1. Perception: understanding environments

2. Decisioning: analyzing and solving problems

3. Actioning: managing and executing tasks

4. Agency: operating independently of human input

5. Adaptability: adjusting to changes

6. Knowledge: applying insights in dynamic contexts
   
   

Each dimension advances along a maturity curve, from simple and reactive at the low end to strategic and independent at the high end.

That progression defines the path from traditional assistants to true AI agents — systems capable of orchestrating actions across multiple systems and contexts, without requiring constant human supervision.

Why this matters for insurance — and other complex service industries

For insurance, the implications are immediate and transformative. Most carriers are still anchored in fragmented conversational experiences: chatbots that answer FAQs, or digital assistants that handle simple policy inquiries. But these minimal-capability systems don’t address the real friction in the insurance value chain.

• Underwriting: AI agents could synthesize application data, external risk factors, and historical policy data to recommend coverage decisions or flag inconsistencies, reducing manual review time and improving accuracy.

• Claims management: Instead of waiting for human triggers, an agentic system could autonomously request missing documents, check claim completeness, and even proactively schedule adjuster visits based on claim type and region.

• Customer service: Rather than handing off to human reps after gathering intent, an agent could complete the task — filing a claim, updating a policy, or negotiating renewal terms in real time.
  
  

These are not hypothetical capabilities—they represent the next step in the agentic capability continuum Gartner outlined. And the insurance sector, with its process-heavy workflows and deep data stores, is a natural fit for this transition.

Don’t underestimate the iceberg: What reliable AI agents require

Gartner’s now-famous “Agent Iceberg” slide drove home a sobering truth: what you see is only the tip.

Above the surface, an AI agent might appear to perform effortlessly, handling customer queries, automating back-office workflows, or optimizing processes. But beneath that visible layer lies the complex foundation that makes it all reliable:

• Guardrails and enforced determinism to ensure predictable outcomes

• Compliance supervision and PII protection

• Regression testing, changelogs, and observability for continuous monitoring

• Fine-tuning pipelines, latency mitigation, and secure orchestration for performance at scale
  
  

For insurers, where data privacy, auditability, and regulatory compliance are non-negotiable, this complexity isn’t a deterrent. It’s a blueprint.

Gartner’s message was clear: you can’t just deploy agents—you must engineer for trust.

Beyond pilots: designing for context and measurable ROI

One of the opening keynote’s most revealing statistics: only 1 in 5 AI projects deliver ROI, and just 1 in 20 achieve transformational impact.

The gap, Gartner said, lies in context engineering — the discipline of connecting data, workflows, and systems in ways that give AI a full understanding of its environment. Without context, even the most powerful model becomes a blunt instrument.

For insurance, this context gap is particularly acute. Customer data lives across policy administration systems, CRM platforms, and claims databases. When AI can’t access or interpret that context, it can’t act intelligently.

That’s why Gartner’s research emphasized ecosystem readiness over individual model performance. Enterprises that integrate AI into their operational fabric via APIs, partner ecosystems, and unified governance see the highest return on adoption.

The agentic enterprise: humans and agents, working together

This idea of an “agentic workforce” isn’t theoretical. It’s already showing up in use cases across service-heavy industries:

• AAA Washington now leverages agentic flows for cross-sell and upsell opportunities during service interactions.

• Pelago evolved its AI from a service bot to a proactive, agentic function that manages entire service workflows.

For insurance leaders, this means shifting perspective, from automating tasks to augmenting teams. The goal isn’t to replace underwriters, claims adjusters, or service reps. It’s to give them intelligent co-workers that handle the repetitive work, so humans can focus on empathy, judgment, and strategy.

From Cool Vendor to category leader

Parloa’s recognition as a Gartner Cool Vendor underscores this very evolution. Where many conversational AI providers remain stuck in reactive mode, Parloa is already advancing the agentic frontier, building systems that integrate perception, decisioning, and action within complex enterprise ecosystems.

By focusing on contextual orchestration and dynamic decisioning, Parloa bridges the gap between LLM power and enterprise reliability—the exact challenge Gartner’s Agent Iceberg warns about.

For insurers, that translates into measurable impact: faster claim cycles, more personalized service experiences, and reduced administrative burden—all built on a foundation of compliance and control.

The takeaway: From talking to truly doing

Gartner’s closing sessions left CIOs with a clear directive: stop measuring AI by what it can say, and start measuring it by what it can do.

Agentic AI is the next competitive edge—not because it replaces people, but because it enables them. And as Gartner’s frameworks show, the enterprises that invest in agentic foundations today will be the ones shaping the market tomorrow.

In short:

AI assistants talk.

AI agents act.

Agentic enterprises lead.