Chloe Maraina is a professional who sees the hidden narratives within vast, complex datasets, viewing big data not just as a collection of numbers but as a canvas for strategic storytelling. With a deep-rooted passion for business intelligence and data science, she has spent years helping enterprises navigate the treacherous waters of digital transformation, focusing on how legacy systems can evolve to meet the demands of a modern, AI-driven world. In our conversation today, we dive into the seismic shifts occurring within the enterprise resource planning landscape, specifically examining how global giants are attempting to balance the security of their most valuable data with the urgent need for openness. We explore the tension between maintaining a “monolithic” control over business processes and the reality of a fragmented AI ecosystem where partners and competitors alike are vying for a seat at the table.
Our discussion centers on the evolution of data infrastructure and the high-stakes gamble of opening up decades of proprietary information to third-party applications. We delve into the concept of the “autonomous enterprise,” the strategic maneuvers involving multi-billion dollar acquisitions, and the technical innovations like zero-copy sharing that aim to eliminate the friction of data movement. Throughout the interview, the overarching theme is one of radical transition—a shift described by industry veterans as being more significant than the advent of the internet itself—where the goal is to transform stagnant repositories into fluid, intelligent assets without losing the governance and compliance that made them indispensable in the first place.
The enterprise landscape is currently witnessing a historic pivot as long-standing platforms attempt to open up decades of legacy data to third-party AI; how do you see this tension between protecting the “crown jewels” and fostering an open ecosystem playing out?
It is a fascinating balancing act that reminds me of the political shifts of the late 20th century, where trying to save a system through reform can sometimes lead to its complete dissolution if not handled with surgical precision. For 53 years, these massive ERP systems have operated as closed, monolithic fortresses, housing the logic and history of roughly 80% of the world’s commercial transactions. Now, the pressure to stay atop the AI wave means these vendors have to offer an unprecedented degree of openness to that wealth of data, which is essentially their biggest competitive bulwark. If they tighten the grip too much by creating “data toll roads” or restrictive API policies, they risk alienating a customer base that is desperate to use third-party AI tools. However, if they open the gates too wide without a clear structural plan, they might lose their identity as the primary portal for enterprise applications, as partners and competitors start jockeying to manage the governance and orchestration layers themselves.
In an era where “agentic AI” is becoming the new buzzword, what are the specific risks and rewards for a major vendor when they encourage partners to disrupt their own established business models?
The reward is survival, plain and simple, because it is far better to have your own partners take a slice of your business with innovative agentic applications than to lose that business to an entirely outside entity. We saw this reality reflected in the recent 100-million-euro investment aimed specifically at bolstering the AI partner ecosystem, a move that signals a desperate need to show that innovation is actually happening on these legacy platforms. The risk, of course, is a repeat of past missteps where a rush to keep up led to “half-baked” solutions, like when developers were reportedly holed up for two weeks in Walldorf and Palo Alto to churn out agents that the market hasn’t fully vetted yet. There is a palpable anxiety that if the “functional stack”—which includes decades of industry-specific know-how and compliance as a service—isn’t seamlessly integrated into these new agents, the vendor becomes just another data source rather than the brain of the operation. It is a high-stakes game where you have to let your partners “disrupt” you just enough to keep the platform relevant, but not so much that you become an invisible commodity in the background.
With the massive $55 to $60 billion spent on SaaS acquisitions over the last decade, how does a company begin to harmonize such a fragmented data landscape to make it “AI-ready” for the modern enterprise?
The sheer scale of that spending spree created a massive “data chaos” problem because each of those acquired companies, from Ariba to SuccessFactors, came with its own unique data models and silos. To solve this, the focus has shifted toward creating a unified Business Data Cloud that acts as a harmonizing layer, combining the organic ERP heritage with these newer cloud-based assets. One of the most critical components of this strategy is the development of prepackaged “data products” that bundle metadata, algorithms, and business records like sales orders into shareable assets. It’s an attempt to turn a decade of expensive acquisitions into a singular, sophisticated data model that can support 26 different industries without requiring a total overhaul of the underlying code. The goal is to make the data feel local and accessible to the user, regardless of whether it originated in a legacy on-premise system or a modern SaaS application, effectively erasing the boundaries that billions of dollars in M&A activity originally created.
Technical innovations like “zero-copy sharing” are often touted as the silver bullet for data integration, but how does this actually change the day-to-day reality for a business trying to leverage its data in environments like Snowflake or Databricks?
Zero-copy sharing is a genuine game-changer because it moves us away from the archaic and risky practice of physically moving or duplicating massive datasets across different environments. Instead of creating a dozen different versions of a single truth, you allow access to the data as if it were local to a third-party environment, much like how you might collaborate on a shared document in a cloud server. This removes the “friction” of data movement, which has historically been the biggest barrier to real-time AI analysis, allowing high-performance tools like Snowflake to query the data where it lives. By moving in-memory database technologies like HANA into this shared cloud environment, enterprises can finally use lightweight transactional records to fuel their AI agents without the heavy lifting of traditional replication. It’s a sensory shift for the data architect; instead of the “heavy lifting” of data migrations, the work becomes about managing permissions and “opening windows” into the existing data stores.
The concept of the “autonomous enterprise” has been met with both excitement and skepticism; why do you think there is such a disconnect between the corporate branding of AI and the practical needs of the workforce?
The phrase “autonomous enterprise” carries a certain cold, mechanical weight that can feel threatening to a workforce already worried about being replaced by technology, which is why some analysts have called it a poor choice of messaging. For the people on the ground, the goal isn’t to create a company that runs without humans, but rather to automate the specific, tedious processes that make sense for their unique workflow. There is a clear gap between the theoretical use cases presented at major conferences and the actual, measurable ROI that a business can see on its balance sheet today. Much of the value we see right now actually comes from basic digital transformation and workflow integration rather than the “magic” of the AI itself, making it very difficult to separate the hype from the functional reality. People want tools that feel like “partners,” like a role-based assistant that helps them navigate a complex procurement task, rather than an overarching system that tries to operate the entire business autonomously from the top down.
What is your forecast for the role of legacy ERP data in the next five years as AI agents become the primary interface for business applications?
I believe we are entering a period where the traditional “user interface” of the ERP system will effectively disappear for the average employee, replaced entirely by agentic assistants that treat legacy data as a foundational “truth” layer. We will see a shift where the value of a platform is judged solely on how well its data can be orchestrated by outside AI tools like Anthropic’s Claude or Google’s agents, rather than the features of the platform itself. The companies that thrive will be those that successfully transitioned from being “closed systems of record” to “open systems of intelligence,” where 80% of the world’s transactions are not just stored, but are actively feeding a global network of autonomous agents. However, this will require a radical cultural shift away from protectionism; if the major players continue to set up “data toll roads,” they will find themselves bypassed by more agile competitors who understand that in the AI age, the flow of information is more valuable than the storage of it. The “monolith” must become a “mesh,” or it will eventually become a museum piece of early 21st-century commerce.
