Peripheral Automation in manufacturing: adding AI and automation without replacing your core ERP
Manufacturing leaders are caught between two pressures, which the consulting industry usually treats as a single choice. Production teams want AI-driven analytics, predictive maintenance, smarter scheduling, and automated workflows. The board does not want to fund another multi-year ERP replacement program with a real chance of missing its business case. Treated as one decision, the two pressures cancel each other out. Treated as separate layers of the enterprise architecture, they resolve into a clear path. That path is Peripheral Automation. The principle is straightforward: extend stable core systems with adaptive data, process, and AI-led capabilities at the periphery, rather than ripping out the core to install a newer version of the same core. Why manufacturing ERP replacement projects keep failing ERP replacement projects fail at a rate the manufacturing industry has learned to tolerate. Gartner predicts that more than 70% of recently implemented ERP initiatives will fail to fully meet their original business case goals by 2027, with as many as 25% failing catastrophically. The cost is not just license fees and the system integrator bill. The real cost is the two to three years the operations team spends in implementation meetings instead of running the business. The pattern repeats because the core ERP carries decades of process logic, master data, and operational habits that are extraordinarily hard to move. Customizations made fifteen years ago by people who have since left the company are still doing useful work. Tear that out, and the replacement project ends up rebuilding the same logic in a new system, slower and at greater cost than planned. The smarter question is not “which ERP should we replace this one with?” but what work the core ERP needs to keep doing well, and what work should sit outside it. What is peripheral automation in manufacturing Peripheral Automation is an enterprise architecture approach that separates the manufacturing technology estate into three layers and confines change to the outer two. The core stays stable. The periphery moves fast. The core: data integrity and transactional record The core is the ERP, the financial general ledger, the MES system of record, and the master data that runs them. The core’s job is to be reliable, auditable, and slow to change. Replacing it is expensive and risky, and most of the time, it is not what the business actually needs. Peripheral Automation keeps the core in place and protects it. The process layer: workflows, automation, and AI agents The process layer is where most manufacturing transformation value lives. Workflow automation in Power Platform, AI agents for triage and decision support, RPA bots for invoice processing, and no-code apps for shop floor data capture all sit on top of the core. The process layer reads from and writes to the core, but is not the core, and the whole layer can be replaced, extended, or retired without touching the ERP. The experience layer: dashboards, interfaces, and conversational AI The experience layer is what plant operators, engineers, finance teams, and executives actually see. Power BI dashboards, Copilot-style assistants, embedded analytics inside Dynamics 365, and mobile field apps all live here. Changing the experience layer does not require a project plan; it requires a sprint. The architectural discipline is that change at the experience and process layers does not propagate into the core. Conversely, the core can eventually be modernized when business needs justify it, without forcing the periphery to be rebuilt at the same time. How peripheral automation adds AI to manufacturing without replacing the core ERP Peripheral Automation in practice means a portfolio of small, additive moves that compound. Four patterns appear across most manufacturing engagements. AI agents on top of the ERP, not inside it An AI agent that reads supplier invoices, validates them against the purchase order, flags anomalies, and posts the clean ones into the ERP delivers most of the value of an “AI-native ERP” without the cost of replacing the ERP. The agent reads ERP data through standard interfaces; the ERP stays untouched. Real-time analytics layered onto historian and MES data Power BI on a Microsoft Fabric data layer reads from the historian, MES, and ERP and presents real-time OEE, yield, and quality views without changing any source system. The plant analytics team builds new reports in weeks. Process automation in the gaps the ERP does not cover Most manufacturers have a dozen processes the ERP could handle in theory but does not handle well in practice: rework approvals, deviation tracking, supplier quality returns, SOP acknowledgments. Power Platform apps and Power Automate flows handle these natively, with a cleaner user experience, and write the necessary records back to the ERP. Predictive models that recommend, not control Predictive maintenance and demand forecasting models trained on production data produce recommendations that surface in the existing scheduling and planning tools. The ERP’s planning engine still runs the plan; AI improves the inputs. Risk stays contained because the model never directly controls the line. How to plan a peripheral automation roadmap for manufacturing A Peripheral Automation roadmap is sequenced by business impact, not by technology. Four principles separate the deployments that compound from the ones that stall. Start with one high-volume, low-risk process The first deployment should be a process that runs many times a day, has a clear measurable cost, and does not affect financial close. Invoice processing, purchase requisition routing, and quality non-conformance workflows are typical first wins. Treat the integration with the ERP as the architecture decision The hardest part of Peripheral Automation is not the AI model or the no-code app; it is the read-and-write interface to the core ERP. Investing in a clean integration pattern, ideally on Power Platform or Microsoft Fabric, pays back across every subsequent project. Sequence by data dependency, not by technology AI and automation depend on clean, contextualized data. Building the data layer first, even minimally, is what makes the rest of the roadmap deliver. A plant with unified, ISA-95-enriched production data deploys a new use case in weeks; a plant without
Manish Godha discusses Peripheral Automation at AI Summit NY
At the AI Summit NY, Manish Godha introduced Peripheral Automation, a novel approach to innovation that integrates cutting-edge technologies like AI and cloud computing into businesses without disrupting core operations. In a dialogue with Romi Mahajan-CEO Exofusion, they explored how Peripheral Automation enables targeted, low-risk experimentation, balancing the need for innovation with business continuity. This human-centric framework emphasizes enhancing customer experiences and operational efficiency while maintaining stability, making it a practical and scalable model for enterprises navigating AI adoption. The launch of PeripheralAutomation.org and the Peripheral Automation consortium further highlights its potential to drive collaboration and refine this transformative approach. Here are some of the interview highlights: Romi Mahajan:Peripheral automation as an entry point to AI—let’s start there. The goal of this discussion is to create a dialogue, so people can better understand how to think about this approach and its applications.Manish, let’s begin with the basics. Tell us about Peripheral Automation and what it means to you as a business innovator. Manish Godha:Peripheral Automation is a concept that integrates contemporary technologies—like AI, cloud computing, and highly specialized SaaS applications—into business operations in a way that aligns with existing business models. Our approach considers the core elements of a business model: what you do, how you do it, and who your stakeholders are—customers, employees, suppliers, and partners. From an enterprise systems perspective, we think of this in layers: These layers help businesses innovate while maintaining operational continuity. Enterprises today use various technologies simultaneously, and they want to innovate quickly. The challenge is doing so without disrupting their existing systems. That’s where Peripheral Automation fits in—it allows targeted innovation without breaking the core. Romi Mahajan:That makes sense. Let’s dig into the dualism you mentioned—disruption versus continuity. While disruption fuels innovation, businesses still need to run efficiently. It’s not about stopping the plane to redesign it mid-flight. How does Peripheral Automation navigate this balance? Manish Godha:Peripheral Automation is rooted in what I call “differential innovation.” Businesses can’t overhaul everything at once—it’s neither practical nor necessary. Instead, you focus on specific areas where innovation will have the most impact. By thinking of the organization in terms of its various units and layers, it becomes easier to identify high-impact opportunities. You innovate within a controlled scope, ensuring the surrounding systems remain stable. This way, you disrupt only what needs to change while the rest of the business continues seamlessly. Romi Mahajan:When it comes to AI and technology adoption, many people think of it as purely a technical issue—“a silicon problem.” But the truth is, it’s often about people and processes. How does Peripheral Automation address these softer, human aspects of AI adoption? Manish Godha:It starts with the business model itself, which revolves around people—customers, employees, suppliers, and partners. A business is most innovative at its interfaces with these people. That’s why the experience layer is so crucial—it’s where differentiation happens. Two businesses might share the same core systems or processes, like invoicing or procurement, but their customer experiences could be worlds apart. By focusing on the experience layer and aligning it with people’s needs, Peripheral Automation fosters innovation that is both meaningful and practical. Romi Mahajan:We’ve seen many headlines about companies that struggle with AI adoption. Some dive straight into large-scale implementations, only to face backlash—whether from customers receiving poor responses or from employees dealing with ineffective tools. Are these failures examples of businesses bypassing the Peripheral Automation approach? Manish Godha:Absolutely. Many of these failures stem from deploying AI wholesale, disrupting core operations in the quest for rapid innovation. Peripheral Automation takes the opposite approach. Instead of automating entire verticals, it identifies smaller, low-risk opportunities for experimentation. These are areas where innovation can be tested incrementally, with backup systems in place to de-risk the process. This method is not only safer but also more cost-effective. You don’t need to build entirely new models from scratch—you refine and scale improvements as they prove successful. Romi Mahajan:That incremental, stepwise process resonates. In a world where AI is often overhyped, real adoption in enterprises is usually much more sober and methodical. That brings us to an exciting announcement you wanted to share. Can you tell us more? Manish Godha:Yes, I’m thrilled to announce the launch of PeripheralAutomation.org. This initiative brings together leading companies—like Advaiya, Exofusion, Nexus Technology, and others—that have extensive experience in innovation and technology implementation. These organizations are pooling their expertise to develop a comprehensive Peripheral Automation framework. PeripheralAutomation.org is live now. The goal is to create a robust, open-source model that benefits businesses across industries. Romi Mahajan:That’s fantastic. So, to anyone listening, head over to PeripheralAutomation.org to learn more about this innovative approach. If you’re interested in contributing or getting your organization involved, be sure to reach out.