Today, we’re speaking with Dominic Jainy, an IT professional whose expertise lies at the intersection of artificial intelligence, machine learning, and enterprise systems. We’ll be exploring one of the most critical challenges facing modern businesses: securely and effectively connecting AI to the core of their operations, the ERP. Our conversation will focus on three key pillars for a successful integration: ensuring AI fits into a company’s existing tech stack without compromise, building a security framework that prevents data leaks by design, and empowering AI to move beyond simple queries to take meaningful, time-saving actions.
The text states that businesses shouldn’t have to adapt their tech stack for AI. How does your solution connect diverse systems, from a Microsoft Dynamics ERP to any SQL database, with a client’s preferred LLM like ChatGPT or Claude? What does that initial setup process look like?
That’s the fundamental principle we operate on. Forcing a business to rip and replace their trusted systems is a non-starter. Our approach is to act as a universal, secure bridge. Whether your financial data lives in NetSuite, your operations run on Microsoft Dynamics 365 Business Central, or you have decades of legacy information sitting in a SQL database, we can connect to it. The beauty is in the flexibility; you’re not locked into a single ecosystem. The initial setup is designed to be surprisingly straightforward. Using our no-code capabilities, the process feels less like a heavy IT project and more like configuring a new app. You essentially point us to your data sources, authenticate securely, and then connect to your preferred large language model, be it Claude, Gemini, or any other. We handle the complex plumbing in the background, ensuring all these disparate systems can talk to each other safely and efficiently.
You emphasize preventing data leaks by design. Can you elaborate on how your access controls differentiate between what a specific user can see versus what the AI agent can see? Please share an example of how these permissions are configured to secure sensitive company data.
This is a crucial distinction and where many AI integrations fall short. Security can’t be an afterthought. Our system is built on a dual-permission model. First, we have the user’s permissions. When a user logs in, the system knows exactly who they are and what data they are authorized to see based on their role. For example, a sales manager for the West Coast region can only see sales data for that territory. Second, we have the AI agent’s permissions, which might be broader to answer complex questions. The magic is how they interact. If that same West Coast sales manager asks the AI, “What is our company’s total revenue for the year?” the AI agent can access all regional data to calculate the total. However, it will only provide the final, aggregated number. If the manager then asks, “Show me the detailed sales from the East Coast,” the system will block the request because it violates that specific user’s permissions, even though the AI itself can see the data. It creates a secure firewall where the AI serves the user without ever over-exposing sensitive information.
The content claims most AI solutions can’t monitor data once it’s accessed. How do your detailed audit trails provide complete visibility and context every time data is read? What specific information gets logged to create a clear accountability trail for every user action?
That lack of visibility is a terrifying prospect for any company. It’s like giving someone keys to the building but having no cameras inside. With Popdock AI, we build a glass house, not a black box. Every single interaction is logged in a detailed audit trail that provides a complete narrative. It’s not just a simple log that says, “User X accessed the customer table at 2:15 PM.” Our logs capture the full context: the specific user who made the request, the exact natural language query they typed, which data the AI accessed to formulate its response, and the final answer provided. This creates an unbreakable chain of accountability. If a sensitive piece of information is ever questioned, you can trace its journey from the database to the user’s screen with complete certainty. This transparency is what gives leadership the confidence to truly embrace AI across the organization.
Moving beyond queries to actions is a key point. Using the example of automatically emailing past-due invoices, could you walk me through the steps to enable an AI agent to perform this task? What kind of measurable time savings have your clients reported from such automations?
This is where the real return on investment kicks in. It’s the difference between having a research assistant and having an actual team member who gets work done. To enable an action like emailing past-due invoices, the process is deliberate and secure. First, an administrator defines the action within the system, granting the AI agent the specific permission to “send past-due invoice emails.” This isn’t a blanket permission to email anyone; it’s a specific, controlled capability. Next, you connect that action to the necessary data—in this case, the ERP data for invoice statuses and customer contact information. Finally, you can set up a trigger, either a manual prompt from a user or an automated schedule. The AI then executes the task, and clients have reported saving hours every week by automating these kinds of repetitive but critical financial communications, freeing up their teams to focus on more strategic work.
What is your forecast for the future of AI integration within ERP systems?
I believe the line between the ERP and the AI will completely dissolve. Right now, we often think of “asking the AI” as a distinct action, like opening a chat window. In the near future, that will feel archaic. AI will become the intelligent, predictive fabric woven directly into the ERP’s interface and workflows. Instead of you asking for a report on supply chain vulnerabilities, your ERP will proactively alert you that a supplier in a specific region is at high risk due to emerging geopolitical events and will have already modeled the impact and suggested three alternative, vetted suppliers. The interaction will shift from being reactive to proactive and predictive. The real power won’t just be in answering questions but in automating complex, cross-departmental decisions before a human even realizes a problem is brewing, all based on a secure, holistic view of the company’s data.