To successfully scale enterprise AI, organizations must focus on enterprise productivity, rather than just individual task augmentation.
While individuals and small groups are realizing productivity gains using Generative AI, two recent studies found that enterprise AI, particularly the use of agentic AI, is not returning value, and many projects never reach fruition.
In particular, Gartner estimates that over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value, or inadequate risk controls.
These were some of the findings in a Gartner report, Emerging Tech: Avoid Agentic AI Failure: Build Success Using Right Use Cases. (The report is only available to Gartner clients.)
“Most agentic AI projects right now are early-stage experiments or proof of concepts that are mostly driven by hype and are often misapplied,” said Anushree Verma, Senior Director Analyst, Gartner, in a press release. “This can blind organizations to the real cost and complexity of deploying AI agents at scale, stalling projects from moving into production. They need to cut through the hype to make careful, strategic decisions about where and how they apply this emerging technology.”
An MIT study had more bad news. Specifically, the GenAI Divide: State of AI in Business 2025 study found that despite $30 billion to $40 billion in enterprise investment into GenAI, 95 percent of organizations are getting zero return. Only 5 percent of integrated AI pilots are extracting millions in value, while the vast majority remain stuck with no measurable P&L impact.
The report is based on a review of over 300 publicly disclosed AI initiatives, interviews with representatives from 52 organizations, and survey responses from 153 senior leaders collected across four major industry conferences.
The report noted that tools like ChatGPT and Copilot are widely adopted. Over 80 percent of organizations have explored or piloted them, and nearly 40 percent report deployment. But these tools primarily enhance individual productivity, not P&L performance.
Meanwhile, enterprise-grade systems are being quietly rejected. Sixty percent of organizations evaluated such tools, but only 20 percent reached the pilot stage, and just 5 percent reached production. Most fail due to brittle workflows, lack of contextual learning, and misalignment with day-to-day operations.
See also: How and Where to Start with AI for Industry
Ensuring Successful AI Scaling Across the Enterprise
AI, and particularly AI agents, are well-suited to industrial environments where complex, variable-rich systems make scripted programming impractical. Agents can parse documentation, identify pertinent information, and integrate it into data models for organized retrieval and decision support.
At the foundation of successful AI agent deployment lies data quality. “Garbage in, garbage out” rings especially true. Organizations must first organize their structured and unstructured data. Otherwise, AI will struggle to deliver value.
A good way to accomplish this is to establish industrial knowledge graphs, contextualizing data across sources so agents “know where to look” for the right information. Even before deploying AI agents, building this data architecture offers immediate benefits: enhanced dashboarding, faster query response times, better troubleshooting, and overall productivity gains across operations.
In addition, there are several things an industrial organization can do to successfully scale its AI agent efforts. They include:
- Begin with Data, Not Models: Organize structured and unstructured data into a well-modeled repository (e.g., a knowledge graph) before layering in AI.
- Prioritize Trust Through Model/Data Pairing: Use well-selected language models in tandem with high-fidelity operational data to support reliable decision-making.
- Provide Precise, Business-Aligned Instructions: Frame agent behavior clearly, just as one would train a human colleague, to ensure consistent, business-focused performance.
In other words, there are three pillars essential to trustworthy agent operations: selecting the right language model (continuously benchmarked for accuracy and reduced hallucination risk); ensuring agents access only vetted, context-rich data stored in a coherent data model; and crafting specific, clear instructions, mirroring the training one would give a new intern, to align agent behavior with business goals. This approach minimizes errors and allows systems to evolve. So, as data models grow richer over time, agent outputs improve without sacrificing reliability.
That ties back to the Gartner findings. “To get real value from agentic AI, organizations must focus on enterprise productivity, rather than just individual task augmentation,” said Verma. “They can start by using AI agents when decisions are needed, automation for routine workflows, and assistants for simple retrieval. It’s about driving business value through cost, quality, speed, and scale.”
