Dominic Jainy stands at the forefront of the next great shift in corporate technology, where artificial intelligence moves beyond simple chat interfaces into the realm of autonomous execution. With a deep background in machine learning and decentralized systems, he has spent years dissecting how complex organizations can bridge the gap between raw data and actionable intelligence. As we navigate 2026, the conversation has shifted from whether AI can write an email to whether it can independently manage a global supply chain or reconcile multi-million dollar accounts. In this discussion, we explore the critical role of “business context”—the invisible web of policies, historical data, and specific configurations that allow an AI agent to function not just as a tool, but as a trusted colleague within a governed enterprise ecosystem.
The following interview explores the transition from generative AI to agentic workflows, emphasizing why generic models often fail in the face of company-specific nuances. We delve into the mechanics of “governed execution,” the necessity of data harmonization across fragmented systems, and the tangible financial impacts seen by early adopters who have moved past the pilot phase into full-scale deployment.
Generic AI agents often understand global accounting standards like IFRS 15 or US GAAP, but they frequently struggle with a firm’s specific internal logic. Why is this local context more valuable than general intelligence when it comes to actual execution?
General intelligence is essentially a map of the world, but local context is the GPS that knows exactly where your specific vehicle is allowed to turn. While an LLM might understand the broad strokes of revenue recognition, it has no inherent way of knowing your company’s unique posting history, document flows, or specific contract nuances. To actually execute work reliably, an agent needs to dive into the granular configurations of your ERP system, understanding the precise calculations that apply to your industry’s regulatory requirements. Without this specific “connective tissue,” an agent is just a well-read observer rather than an active participant. It is the difference between a consultant who knows the theory and a veteran employee who knows exactly which exceptions apply to a long-standing client’s account.
How does the transition from an agent that merely suggests an action to one that actually executes a transaction change the way leadership views risk and governance?
We are moving away from a world where an AI says, “I think you should do X,” and into a reality where it says, “I am authorized to do X, and I know which systems to touch to get it done.” This shift to “governed execution” is profound because it requires the agent to operate within the same strict controls, approval rights, and compliance frameworks that a human would. For a single payment, the agent must track a complex web of objects—the supplier, the specific product, the invoice, and the underlying contract—while ensuring every step is recorded in an auditable workflow. This isn’t just about speed; it’s about building a digital paper trail that is more accurate and transparent than traditional manual processes. When leadership sees that an agent can reconcile a payment under strict policy with a person-in-the-loop for final sign-off, the anxiety around autonomy begins to melt away.
Data is frequently scattered across various applications with different identifiers—one system says “IBM,” another says “International Business Machines.” How does an agent overcome this “identity crisis” to ensure it isn’t making decisions based on fragmented information?
This is where the strategy of “hoping the LLM figures it out” completely fails in an enterprise setting. You cannot rely on luck when managing critical supplier relationships or high-value transactions. We solve this by implementing a context layer above the raw data, utilizing ontologies and knowledge graphs that reconcile these scattered records into a single, trusted profile. This harmonization process ensures that the agent understands that “IBM” and “IBM Inc.” are the same entity, governed by the same global contract. By using dynamic survivorship to weigh which data source is the most recent or trustworthy, we provide the agent with a “semantic ground truth” that allows it to navigate complex relationships without getting lost in the clutter of legacy data.
You’ve noted that many companies are surprised by the reality of their own internal processes. How does process mining reveal the gap between what executives think is happening and what the AI agents actually encounter?
It is a common phenomenon for a leadership team to claim they have one standardized procure-to-pay process, only to find out through tools like Signavio that they actually have 400 different variations running simultaneously. These invisible “shadow processes” are the natural result of regional exceptions, legacy habits, and manual workarounds that have calcified over time. An AI agent cannot be expected to solve a problem if it is constantly hitting these undocumented walls. By tracking how workflows actually run in real-time, we can draw a clear line for the agent: where there is room for creative problem-solving, it can act, but where compliance requires a fixed path, the agent follows a rigid, non-negotiable workflow. This transparency allows organizations to finally clean up their operational inefficiencies before automating them.
When we look at the bottom line, can you share how this shift toward agentic workflows is translating into measurable financial success or time recovery for global organizations?
The impact is no longer theoretical; we are seeing massive gains in both efficiency and top-line growth. For instance, at PwC, a complex billing task that used to consume 35 minutes of human effort has been compressed down to just five minutes, all while increasing the accuracy of the output. In the heavy machinery sector, Daimler Trucks North America utilized these governed workflows to transform their fleet contract bidding. They saw their win rate skyrocket from roughly 10% to more than 40%, which generated a staggering $70 million in financial impact over a single 12-month period. These numbers prove that when agents are grounded in real business context, they don’t just save time—they fundamentally change the competitive posture of the company.
Logistics is a field defined by chaos and constant exceptions. How do specialized agents handle something like an inventory shortage differently than a human team facing a traditional “fire drill”?
In a traditional logistics fire drill, a team might spend hours jumping between spreadsheets and different software modules to find an alternative source for a missing component. A specialized agent, however, can look across all warehouses instantly, checking which inventory is already committed to other orders and which expiration rules apply to the stock on hand. Instead of a multi-step scavenger hunt, the process becomes a “one-click” resolution where the agent identifies the best alternative transportation and prepares the necessary approval requests. We’ve seen customers in the natural ingredients industry recover about a third of their team’s time by letting agents orchestrate these logistics exceptions. It turns a high-stress crisis into a routine, governed decision-making process.
What is your forecast for the evolution of these autonomous agents as they begin to communicate across different vendor boundaries?
From 2026 to 2028, the biggest hurdle won’t be internal automation, but how agents from different technology stacks—like SAP and Microsoft—talk to one another without losing the “business meaning” of the data. We are already seeing the emergence of protocols like Agent2Agent (A2A) which allow these assistants to coordinate around a business objective rather than being confined to a single piece of software. My forecast is that we will move toward a “borderless” enterprise context, where the policies, relationships, and controls that make an action trustworthy travel with the work, regardless of which system is being used. Interoperability will shift from being a technical integration headache to a strategic necessity, as agents begin to negotiate and execute transactions across the entire global value chain.