51% of organizations rely primarily on public AI tools such as ChatGPT and Microsoft Copilot, while only 20% report enterprise-wide AI deployments built on governed AI frameworks. At the same time, more than 80% of mid-market and enterprise organizations are actively launching AI and machine learning initiatives, and 62% identify AI as a strategic business priority. Yet despite this momentum, many organizations remain unprepared for production-scale AI.
In this episode of AppDevANGLE, I spoke with Susan Laine, Chief Data Technologist at Quest Software, about the widening gap between AI ambition and enterprise readiness. Rather than focusing on larger models or faster infrastructure, the discussion centered on the foundational challenges of data governance, semantic context, production operations, and AI trust.
The conversation reinforced an important reality: enterprise AI is no longer constrained by model capability. It is increasingly constrained by the quality, governance, and usability of enterprise data.
The AI Governance Gap Is Growing Faster Than AI Adoption
Many organizations have made remarkable progress adopting AI tools. What has not kept pace is governance. As Laine explained, earlier AI initiatives primarily required organizations to govern how humans interacted with data. Agentic AI fundamentally changes that equation because the consumer of enterprise data is increasingly software itself.
“These data discrepancies are going to scale faster and further than ever before… A human can run a handful of queries a day… An agent issues orders of magnitude more requests, and they don’t have to apply this human judgment.”
That distinction matters. Business users naturally identify inconsistencies, question unexpected results, and collaborate to resolve conflicting definitions. AI agents do not. They operate at machine speed, consuming whatever data they receive without applying organizational context unless that context has been deliberately encoded.
As enterprises deploy autonomous agents across customer service, operations, analytics, and software development, poor governance no longer creates isolated reporting errors. It creates the potential for those errors to be amplified continuously across entire business processes. The challenge now is governing how AI systems interpret and act on data.
Data Readiness Remains the Hidden AI Bottleneck
Much of today’s AI conversation focuses on foundation models, GPUs, and inference infrastructure, but Laine argues that organizations often overlook a more fundamental problem. “AI is going to find those dirty socks underneath the bed quite easily.”
Development environments often perform well because AI models are trained and validated using curated datasets. Production environments tell a different story. Once AI systems begin interacting with live operational data, inconsistent definitions, poor metadata, duplicate records, and low-quality data quickly become visible. Models that performed well during testing begin producing inconsistent or unreliable outcomes—not because the models changed, but because the underlying data changed.
This helps explain why many AI pilots struggle to scale successfully. Organizations frequently believe they have an AI deployment challenge when, in reality, they have a data quality challenge. The conversation increasingly shifts from model optimization toward establishing trusted data foundations that AI systems can consistently rely upon.
Context Is Becoming an Enterprise Performance Multiplier
One of the most practical discussions centered on semantic context. Enterprise AI often relies on large prompts that repeatedly explain business definitions, relationships, and organizational terminology before a model can answer relatively simple business questions.
Laine explained that semantic layers significantly reduce this overhead. “The more context and the more semantics that you’re using, the better the answers are.” Rather than forcing AI systems to repeatedly infer concepts such as customer profitability, revenue definitions, or business hierarchies from lengthy prompts, semantic layers allow those concepts to exist as reusable business knowledge.
The impact extends beyond answer quality. Organizations also reduce token consumption, shorten prompts, decrease inference costs, and improve consistency across AI applications. As AI usage expands across thousands or millions of prompts, these operational efficiencies compound quickly. Context is no longer simply about improving accuracy. It is becoming an economic advantage.
AI Governance Must Scale Alongside AI Development
Another important theme was the evolution of governance itself. Historically, governance processes often slowed application delivery because documentation, modeling, and metadata creation required extensive manual effort.
Laine described how many organizations are now applying AI to accelerate governance rather than bypass it. “The vendors and the clients are starting to use AI for AI.”
Instead of eliminating governance activities, organizations are using AI to automate logical data modeling, identify trusted data sources, generate metadata, create data products, and apply governance controls more efficiently.
This represents a meaningful shift. AI is becoming part of the governance process itself. Rather than choosing between innovation speed and governance discipline, enterprises are increasingly using AI to achieve both simultaneously. This approach becomes particularly important as citizen developers and line-of-business teams build AI-powered applications outside traditional data management organizations. Governance cannot remain a manual process if AI development becomes increasingly automated.
AI Adoption Is Outpacing Organizational Readiness
Perhaps the strongest observation from the discussion centered on organizational maturity. Laine described what she calls the “AI adoption gap,” where “AI is advancing much faster and becoming much more mature than our capacity to keep up with it from an adoption perspective.”
Technology continues to advance rapidly. Organizational processes, governance structures, ownership models, and operational practices evolve much more slowly. That disconnect increasingly explains why many enterprises struggle to move beyond successful pilots into repeatable production deployments.
The challenge is no longer convincing organizations to invest in AI. The challenge is helping organizations build the operational maturity necessary to use AI responsibly at enterprise scale.
Analyst Take
Enterprise AI discussions have largely shifted beyond model performance. The organizations creating sustainable business value are increasingly distinguished by something far less visible: their data foundations.
Susan Laine makes a compelling point that governance is no longer simply about regulatory compliance or data stewardship. As AI agents begin consuming enterprise data autonomously, governance becomes an operational prerequisite for trustworthy AI.
Equally important is the role of semantic context. Many organizations continue treating prompts as the primary interface between humans and AI. Over time, I expect semantic models, business ontologies, and governed metadata to become the reusable enterprise context layer that enables AI systems to operate efficiently, consistently, and economically.
The most successful AI programs over the next several years will likely invest less time debating model selection and more time improving data quality, semantic consistency, governance automation, and operational trust. Enterprise AI success increasingly depends on preparing data before deploying models. Organizations that recognize that shift today will be significantly better positioned to scale AI tomorrow.
