With over 15 years of experience navigating the complexities of enterprise resource planning, Dominic Jainy has witnessed firsthand how emerging technologies reshape the factory floor. As an IT professional specializing in the intersection of artificial intelligence and blockchain, he focuses on the critical infrastructure required to make these innovations functional rather than just theoretical. In this discussion, we explore the shift toward AI-driven manufacturing within the Dynamics 365 ecosystem, examining why the success of advanced automation depends less on complex algorithms and more on the integrity of the underlying functional data and the expertise of the people who manage it.
The conversation covers the strategic necessity of clean data masters, the specific staffing hurdles manufacturers face when bridging the skills gap, and the phased implementation strategies that prevent AI from amplifying existing operational errors.
Many manufacturers find that AI agents often fail when layered on shaky ERP foundations. How do you audit item masters and routings to ensure accuracy, and what specific data points are most critical for a sourcing agent to function without making bad decisions?
To ensure an AI agent doesn’t start making expensive mistakes, we have to look at the audit process as a foundational cleanup of the D365 environment. We start by validating that the item masters aren’t just filled out, but are actually reflective of current shop floor realities, ensuring that every Bill of Materials (BOM) and routing is synchronized with physical production. For a sourcing agent to provide any real value, it requires a “clean” diet of specific data points: accurate purchase agreements, up-to-date vendor masters, and real-time supplier scorecards that track delivery timelines and quality metrics. Without these verified inputs, the agent will confidently select the wrong supplier based on outdated lead times, essentially accelerating a bad decision-making process.
Large enterprises are adopting AI at rates exceeding 75%, often using digital twins or real-time costing monitors. Could you walk through a scenario where a production costing agent flags a cycle time drift, and what steps a team should take to validate that shop floor data?
In a high-pressure environment, a production costing agent might detect that a specific assembly line is taking 15% longer than the standard time defined in the D365 routing. The immediate reaction shouldn’t be to blame the AI, but rather to use that flag as a diagnostic tool to investigate whether an operator picked up the wrong material or if there is a mechanical bottleneck. The team must validate this by comparing the agent’s real-time execution data against the historical standard costs and the physical output recorded in the ERP. It is a sensory process where the digital twin alerts you to a pulse change, but the human team must verify if the “fever” is caused by a data entry error or a genuine production floor inefficiency.
There is often a significant skills gap when bridging ERP configuration with AI readiness. Why should a company prioritize hiring a senior functional consultant over a data scientist initially, and how does this choice impact the long-term reliability of automated purchase order workflows?
The reason to prioritize a senior FSCM functional consultant is that AI agents are consumers of data, not creators of it; they cannot fix a broken configuration. A data scientist can build a brilliant model, but if the underlying Dynamics 365 setup doesn’t reflect how the shop floor actually operates, the model is useless. By bringing in a consultant to clean up item masters and validate configurations first, you ensure that the automated purchase order workflows are pulling from a single version of the truth. This functional groundwork is the unglamorous prerequisite that prevents your automated systems from spiraling into chaos six months down the line.
The most successful implementations follow a “crawl, walk, run” approach, starting with modest automation targets like 10% of purchase decisions. How do you select the first high-value process for automation, and what metrics do you use to build organizational confidence during these early phases?
Selection usually begins with a high-volume, repeatable process where the data is already relatively stable, such as procurement or warehouse slotting optimization. We look for a “win” that is visible but low-risk, such as aiming to automate just 10% of purchase order decisions to prove the logic works. When we hit a metric like 12% automation with zero errors, it builds a tangible sense of relief and confidence among the staff who were previously skeptical. This incremental success allows the organization to scale their expertise—perhaps using a contractor for the data audit phase and a Copilot specialist later—without the overwhelming pressure of a total system overhaul.
With ERP implementation failure rates reaching up to 75%, adding AI agents can sometimes amplify existing errors. What are the common warning signs that an AI agent is consuming “dirty” data, and how do you restructure D365 configurations to prevent these automated mistakes?
The most glaring warning sign is “confident hallucination,” where the agent suggests procurement volumes or costing adjustments that defy the physical reality of your inventory levels or budget. If you see your sourcing agent ignoring landed costs or quality scorecards that you know are critical, it means your D365 configuration is likely scattered across disconnected systems. To prevent this, we restructure the environment to centralize all vendor and production data within Microsoft Fabric or the core FSCM modules, ensuring there are no “dark corners” of data. We have to treat the ERP as the brain and the AI as the hands; if the brain’s memory is faulty, the hands will inevitably perform the wrong tasks.
What is your forecast for AI agents in manufacturing?
I forecast that within the next three years, the role of the traditional D365 functional consultant will evolve into that of an “AI Orchestrator.” We will see a shift where 76% of large enterprises move beyond simple pilots to fully integrated digital twins that manage the entire supply chain autonomously. However, the “AI skills gap” will remain the single biggest barrier, meaning that the manufacturers who win won’t necessarily have the best algorithms, but will have the cleanest data and the most adaptable functional teams. The human element will actually become more critical, not less, as we move from manual data entry to high-level oversight of these automated agents.