Humans matter, AI still in flux: Lessons from Gartner summit

Gartner analysts report that 90% of AI pilots fail to reach production, citing a lack of human-centric governance and workflow fit as primary causes. At the Application Innovation & Business Solutions Summit in Las Vegas, experts stated that human ingenuity remains the singular compounding factor for scaling AI capability rather than inconsistency.

Why do most AI pilots fail to reach production?

Only 5% of AI pilots currently reach production, according to data presented by Gartner. Senior analysts Jason Wong and Brent Stewart attribute this collapse in business value to organizations treating AI as an additive tool rather than a multiplicative system.

Stewart and Wong define the equation for AI business value as: (model capability x workflow fit x trust x governance) to the power of human ingenuity. If any single variable within the parentheses equals zero, the total value of the project becomes zero.

This means optimizing for the most powerful model is useless if the workflow fit is poor or if employees lack trust in the system. Organizations often focus on model capability while neglecting governance, leading to projects that never leave the prototype phase.

How does the “human-in-the-loop” prevent AI failure?

Human ingenuity is the primary amplifier for enterprise AI, according to analysts Jason Wong and Brent Stewart. They argue that AI does not replace human activity but compresses the cost of iteration, allowing humans to focus on business transformation.

Aaron Lord, a Gartner senior director analyst, compares current AI cognitive abilities to those of a child. He notes that while a model may know a tomato is a fruit, it lacks the judgment to know it does not belong in a fruit salad. This gap necessitates human guardians to keep development on track.

Birgi Tamersoy, also a senior director analyst at Gartner, emphasizes that a 99% success rate inherently means a 1% failure rate. Because real-world situations contain too much ambiguity for binary understanding, Tamersoy states that human risk management is vital.

Deepak Seth, a senior director analyst, adds that while AI may outperform humans in low-complexity tasks—such as autonomous driving versus human-caused accidents—AI agents are not a direct replacement for people. The “human-in-the-loop” serves as the final bottleneck to stop errors before they scale.

What is Agentic AI and how will it change workflows?

The industry is shifting from LLM-supported chatbots toward agentic AI. Jithin Bhasker, general vice president at ServiceNow, positions agentic development as the final stage of application development, following low-coding and AI-assisted “vibe coding.”

Deepak Seth predicts agentic AI will eventually handle all basic-level roles where decision complexity is low, specifically in customer service, IT Service Delivery Lifecycle (SDLC), and IT Service Management (ITSM).

Peter Vaccarella, global head of solutions consulting at Camunda, argues that embedding AI into existing systems is insufficient. He states that every current enterprise process is “legacy.” To solve this, Camunda developed ProcessOS, an operating system designed for AI agents to re-engineer workflows from scratch rather than simply augmenting old ones.

How does governance impact the ability to scale AI?

Governance is the primary path to establishing trust, which the Gartner keynote described as the “new first principle of user experience.” Greg Clock of Plat4mation notes that building this trust early is often the deciding factor in securing long-term customer contracts.

Gartner Senior Director Analyst George Sellner warns that organizations skipping governance steps often “scale inconsistency instead of capability.” He argues that fragmented governance across functions makes AI risky and inconsistent.

Jason Wong notes that slow scaling is often a symptom of poor governance rather than a result of it. By integrating governance into the development process from the start, enterprises avoid spending time “putting out fires,” which Sellner says steals time from achieving mission impact.

Comparison: Standard AI vs. Agentic AI

Frequently Asked Questions

What is the main reason AI pilots fail?
According to Gartner, 90% fail because organizations neglect the multiplicative factors of workflow fit, trust, and governance, focusing only on the model’s capability.

Can AI agents replace human employees?
No. Analysts Jason Wong, Brent Stewart, and Deepak Seth emphasize that AI is an amplifier, not a replacement, and that humans must remain in the loop to manage risk and ambiguity.

What is “ProcessOS”?
Developed by Camunda, ProcessOS is an operating system for AI agents designed to re-engineer legacy enterprise workflows rather than just adding AI to existing ones.

Why is governance necessary for AI scaling?
George Sellner of Gartner states that without centralized governance, organizations scale inconsistency and risk rather than actual business capability.

How can companies improve AI adoption?
By treating trust as a core part of the user experience and ensuring governance is built into the system from the beginning, rather than added as a regulation at the end.