Artificial intelligence has dominated headlines for at least the past two years. From generative AI breakthroughs to chatbots that seem almost human, the hype is impossible to escape. But here’s the sobering reality: according to a recent report published by MIT’s Project NANDA, “The GenAI Divide: State of AI in Business 2025”, 95% of generative AI projects fail to deliver the ROI businesses expect.
For CIOs and business leaders, this is a critical moment. AI is transformative — but transformation doesn’t often come easily. The real question is: what separates successful AI adopters from the rest of the pack?
This blog, the first of two, explores why AI projects stall, the shift from digital to agentic transformation, and the foundational role of integration and automation in turning AI hype into business value.
It draws on a LinkedIn live presentation hosted by Patricia Moore (senior manager of innovation programs at Boomi) and features Boomi CEO Steve Lucas and Chris Hallenbeck, Boomi’s senior vice president of AI and Platform. They discuss the MIT study and how to move beyond AI pilot projects to deliver real, measurable ROI with Boomi’s integration and automation platform.
The Harsh Truth: AI ROI Is Rare
The MIT study at the heart of this discussion was no lightweight survey. It included 150 executive interviews and 300 company-level surveys across industries. The result? A significant confirmation that while AI potential is real, execution often lags far behind.
The study identified three main barriers to ROI:
- Integration: AI agents need real-time access to clean, connected enterprise data.
- Workflow Automation: Too many pilots sit in isolation, never embedded into business processes.
- Education: Organizations lack understanding of what’s truly possible with AI.
AI is no longer a chip or model problem. The challenge now lies in connecting data, orchestrating workflows, and making AI practical at scale.
From Digital Transformation To Agentic Transformation
For nearly two decades, “digital transformation” has been the guiding mantra of enterprise IT. At first, it meant converting paper processes into digital form. Later, it came to mean automation — streamlining deterministic, rule-based processes with software.
But deterministic logic has its limits. As Lucas explained:
- Deterministic systems work on if/then/else rules.
- AI, on the other hand, enables probabilistic reasoning. It can make decisions in gray areas where rules fail.
With the dawn of agentic transformation, instead of rigidly coded processes, businesses can now deploy AI agents that:
- Reason through exceptions
- Handle ambiguity
- Learn from historical context
- Adapt to new inputs
Agentic transformation is about rethinking workflows, not just digitizing them. It’s about blending deterministic processes (like payroll) with AI agents that can reason and handle probabilistic tasks like expense approvals to increase agility and efficiency.
Hype vs. Reality: Separating Signal From Noise
Of course, the clamor around AI agents is deafening. Every vendor seems to have “agentic” slapped on their product. But as Hallenbeck observed, only a handful of companies are actually delivering enterprise-grade capabilities.
The hype creates confusion. Is agentic AI just chatbots? A glorified assistant? Or something more?
The reality: true agentic systems are about embedding AI agents into standard operating processes (SOPs), enabling them to access systems securely, gather information contextually, and trigger next steps in workflows.
This is more than the fevered dream of an overwrought marketing team — it’s about reimagining and rebuilding enterprise processes end-to-end.
Why Most Pilots Stall
So why do so many AI pilots fail? A few recurring stumbling blocks:
- Custom coding overload: Building agents from scratch takes weeks of Python, APIs, and security layers. By the time it’s working, it’s obsolete.
- Scalability gaps: Even successful proofs of concept often can’t handle enterprise-scale deployment.
- Security blind spots: Pilots ignore enterprise requirements like prompt injection defenses, encryption, and audit trails.
(A prompt injection attack is a GenAI security threat where an attacker deliberately crafts and inputs deceptive text into a large language model (LLM).)
The result? Companies remain stuck in pilot purgatory — impressed by AI demos but unable to deploy them in production.
The Way Forward
Agentic transformation requires:
- Platforms focused on integration that links AI with real-time, quality data.
- Workflow automation to embed AI into business processes.
- Enterprise-class security and governance to support scale and regulation.
These aren’t optional “nice-to-haves”; they’re the foundation for turning pilots into production.
In the next blog, we’ll explore how Boomi addresses these needs with its end-to-end agentic platform and examine some real-world customer use cases.
You can watch the entire live presentation here on demand. And, as you scale your business, be sure to check out Boomi Agentstudio to strategically balance the risk and rewards of AI agents.