The crowd in Nashville was not a typical enterprise software audience. The 4,100 attendees at Epicor Insights 2026 were mostly customers who have been running Epicor software for a decade or more. VMC Group since 1996. Cornell Pump since its founding. Tilton Group, the 2026 Customer Innovation Award winner, hitting 99.5 percent service levels after migrating to Kinetic Cloud. This is a community that shows up because the software is woven into how their businesses operate, not because they are evaluating alternatives. Understanding that room changes how you read the announcements that came out of it.
This is the third post in the Epicor series. The first covered the vertical AI strategy in November 2025. The second, published May 2, laid out what to watch at Nashville: the cloud migration timeline, outcomes-based agent pricing, and carbon cost roll-up. This one reports what was actually on stage, in the analyst session, and on the expo floor.
What the CEO Said Before the Product Slides
CEO Steve Murphy opened the keynote with two moves that set the tone. First, he apologized for billing problems, promised personal involvement until they were resolved, and offered direct access at a Thursday support session. That is not a typical conference opener. For a room full of customers who have been on the platform long enough to have opinions about support quality, it was the right call.
Second, he gave a five-point AI assessment before touching a single product announcement: AI is powerful and frightening simultaneously; layoffs attributed to AI mostly reflect over-hiring, not automation displacement; AI will create jobs rather than eliminate them net; the knowledge gap between what people write about AI and how it actually works is real; and the forces of competition will continue finding ways to use human creativity. He was direct about hallucination rates, about the difficulty of prompt engineering, and about why the ERP data layer matters as a grounding mechanism for AI outputs.
"The best models hallucinate more than ten percent of the time. Some are closer to ninety. If you hired someone who said they were great at the job but hallucinated between ten and ninety percent of the time, you'd worry about that." — Steve Murphy, keynote
That framing was deliberate. Murphy spent credibility before spending product claims, which made the Cognitive ERP announcement that followed land with more weight than it would have as a cold open.
Cognitive ERP: The Architecture and the Argument
Epicor President Vaibhav Vohra positioned Cognitive ERP as six capabilities sitting above the existing Industry ERP and Data Engine layers: Industry Data Model, Vertical Agents, LUX Agentic Design System, Agent Foundry, Headless ERP, and Time to Value. Epicor claims to be the first ERP company in the world to ship all six simultaneously. That claim invites scrutiny on the definition of "ship," but the architecture underneath it is coherent.
The foundational argument was made on two slides that were more analytically interesting than most of the product demos. The first stated that ontology is the model of the business, not just the database, and showed a network graph of manufacturing objects, Part, JobHead, JobOper, Company, Vendor, Warehouse, linked as a business semantic model rather than a database schema. The second followed: AI without ontology is just guessing at your data. The three properties that follow from grounding AI in that ontology are Grounded (agents reason in the customer's own data, not generic abstractions), Accurate (every answer is traceable to a specific Job, Part, or Operation), and Actionable (actions on objects respect existing rules and approval flows, so AI can act rather than merely advise).
This is the strongest differentiation argument Epicor has made in years. Every horizontal AI vendor selling into manufacturing is working from generic language models that have no intrinsic understanding of what a job order, a routing, or a bill of materials actually means in a discrete manufacturing context. Epicor is betting that vertical ontology, built from decades of transactional data across its customer base, is a durable moat. The bet is testable: either the agents produce fewer errors and require less human correction than generic alternatives, or they do not.
Agent Foundry is the execution layer. Manufacturers can build custom agents, access pre-built vertical agents from a launching Prism Agent Marketplace, or bring external AI through the open-source Model Context Protocol that Epicor has explicitly embraced. The Quoting Intelligence Agent demonstrated live showed a 41-minute quote cycle and 92 percent same-as-except match rate, converting customer drawings into routing instructions, material recommendations, and Kinetic-ready work orders. That is a specific use case with a measurable outcome.
Headless ERP means Kinetic logic and data surfaces through APIs without requiring the standard Kinetic interface. Combined with LUX, the agentic design system that governs how both humans and AI agents interact with Kinetic data, Epicor is positioning the ERP core as infrastructure rather than interface. The Kinetic screen you see can be customized without code; the ERP underneath remains the system of record and governance.
The data ambition went further in the analyst session. Epicor leadership described a World Model, an industry-level dataset assembled from anonymized, opted-in transaction data across the Epicor customer base, designed to surface macro supply chain insights no individual customer could generate from their own data alone. Vaibhav Vohra cited the ability to predict Consumer Price Index movements more accurately and faster than government sources as one example. That claim requires independent verification, but the directional thesis, that an ERP company sitting on decades of supply chain transaction data has a data asset that pure AI vendors do not, is structurally sound.
The On-Prem Question Got an Honest Non-Answer
The analyst session surfaced the question directly: highly regulated industries outside the US are scared of going to the cloud. Is there a bridge, or are you dealing with the extremes?
The Epicor leadership response acknowledged the concern and pivoted to the majority case: most of its customer base are mid-market manufacturers with small IT departments for whom removing infrastructure burden is a net positive. That is accurate for most of the Insights audience. It did not address the harder subset.
What Epicor's cloud documentation confirms is that Kinetic runs on Microsoft Azure and places customer data in the Azure region geographically closest to them. Customers with data residency requirements are encouraged to seek legal advice. Epicor states it will work with customers who have business, legal, or technical preferences to try to accommodate them. "Try to accommodate" is not a sovereignty guarantee.
Three categories of on-prem international customers face meaningfully different situations. Customers with data residency requirements, where data must stay within national borders, may be satisfied by Azure's regional datacenter footprint if Epicor can contractually bind the tenant location. Customers with infrastructure control requirements, which cover defense-adjacent manufacturers, certain financial services suppliers, and some government contractors, face actual legal or contractual bars that a regional Azure deployment does not resolve. Customers whose concern is operational control and change management familiarity, which is the majority of the international on-prem base based on what came through in the analyst discussion, have a different problem that the partner ecosystem is better positioned to solve than Epicor centrally.
The active support runway through 2029 gives every category real time. The mistake is treating that time as a reason to defer the planning conversation rather than as the planning window itself.
The Expo Floor Maps Every Gap
The expo hall partner booths at Insights 2026 are where analyst work gets honest. Five conversations illustrated the full picture of what Epicor's native platform has historically required partners to complete.
| Partner | Functional Gap | Migration Complexity Signal |
|---|---|---|
| BarTender by Seagull | Label and track infrastructure across the supply chain. 100B+ labels annually, 250K+ customers, 175 countries. | Global footprint means this integration exists in virtually every Epicor manufacturing deployment. Re-certification at cloud migration is not optional. |
| Precise Business Solutions | Spend Management via Advanced Requisition Management, Projects and Contracts via Advanced Project Management, Counter Sales via Point of Sale. Epicor Platinum ISV Partner. | Platinum ISV designation means Epicor has formally acknowledged these gaps. All three modules require re-validation post-migration. |
| Octave | Advanced Quality Management System for Epicor Kinetic. Critical in aerospace, automotive supply chain, food and beverage. | Quality management is often the most customized module in any manufacturing ERP. Migration complexity here is high regardless of technical readiness elsewhere. |
| SPARXiQ | Dynamic pricing intelligence for distribution. Win-rate optimization at the quote level, from too-low to too-high with market-rate positioning. | Pricing logic embedded in quoting workflows is among the stickiest customization debt in Kinetic environments. |
| Phocas | Financial planning, rebate management, and CRM analytics via the EDA Suite. Makes data-driven decisions faster above the ERP transactional layer. | Rebate management for distributors handling complex programs across hundreds of vendors does not migrate casually. This is a planned workstream, not a side task. |
Each of these partners exists because Kinetic historically left a functional gap that customers needed filled. The ecosystem is deep, technically integrated, and genuinely valuable to the customers it serves. It is also the hidden migration cost that no Ascend program projection quantifies. A manufacturer who has spent years building a functioning stack around Epicor is not migrating one system. They are migrating six.
The customer who uses BarTender, Phocas, Precise, Octave, and SPARXiQ alongside Kinetic is not moving one ERP to the cloud. They are coordinating six simultaneous migration tracks, each with its own integration testing, re-certification, and business continuity requirement.
The more interesting strategic question is whether Agent Foundry changes this dynamic over time. If the Prism Agent Marketplace becomes the distribution channel for ISV-built agents rather than ISV-built integrations, the ecosystem relationship shifts from gap-filling to capability-extending. That is a different business model for both Epicor and its partners, and it was visible in embryonic form at the Insights expo floor. The conversations at these booths were not defensive. Partners were positioning around what they build on top of Epicor's agentic infrastructure, not just what they compensate for in the current platform.
Customer Evidence That Traveled Well
Two customer stories from the keynote and general sessions deserve mention because they illustrate what the platform actually produces at maturity.
Tilton Group, the 2026 Customer Innovation Award winner, is a family-owned manufacturer that migrated to Kinetic Cloud and built AI-driven forecasting at the core of its operations. The result: 99.5 percent service levels, near-perfect inventory accuracy, and what the keynote described as confident to-the-dollar planning. Tilton's hockey team also won the night before. Both facts landed with the audience.
Cornell Pump went live on Kinetic in two weeks after acquiring a company running NetSuite and using Epicor's AI-assisted database mapping tools to migrate. The IT manager reporting directly to the CFO described it on stage as "insane" and immediately qualified that two weeks would not be standard. The honest qualification made the story more credible, not less. The underlying point, that Ascend tooling compressed a migration that would historically have taken months, is the one Epicor needed to land with an audience full of customers dreading their own migration timelines.
The CEO Champion Award went to a Canadian distributor founded in 1906, now in its 120th year of operation, that centralized operations across 171 locations by migrating to
