Epifor MES AI is bringing agentic AI into factory-wide production planning, giving manufacturers a way to identify late orders, investigate bottlenecks and make planning decisions using information from across the plant rather than looking at individual machines in isolation.
For manufacturers, the important distinction is scope.
Machine-level AI can help optimize a particular process or piece of equipment. Epifor MES AI takes a different approach. It works around the production order, combining information including routings, priorities, delivery dates, capacity, downtime and quality to give production teams a wider view of what is happening and what needs attention.
The platform is particularly aimed at labor-intensive and custom manufacturing environments, including operations where cutting, bending, welding, assembly and quality processes must be coordinated to deliver an order.
That broader context matters because integration remains one of the biggest MES challenges manufacturers face. A 2026 Rockwell Automation study of 1,560 manufacturing and industrial decision-makers found that just 23% reported full integration across ERP, PLM, quality and operational technology systems, while 44% ranked integration as their top MES buying requirement.
We spoke with Giannis Savvaidis, General Manager at Epifor MES, about why the company is introducing agentic AI, how the technology works across different production processes, what it can currently do autonomously and what manufacturers need to consider before introducing it into an existing factory.
Why has Epifor introduced AI into its MES platform?
MTN: You’ve recently introduced AI capabilities into the Epifor platform. What prompted this move, and what manufacturing problems did you specifically want the AI to solve?
Giannis Savvaidis:
The main goal was to add another layer of intelligence.
We wanted to solve problems that previously meant pulling information from different systems, preparing reports manually, and doing work outside those systems in Excel. Making data easier to access was part of it, but the bigger step was helping people combine that information and use it to make decisions.
Manufacturing planning is a perfect example. Every manufacturer that works with Epifor plans differently, because every factory works differently. Rather than imposing one planning method, we wanted manufacturers to bring in their own way of working: how they set priorities, schedule and organize the shop floor.
With our agentic AI, a manufacturer can provide those rules and connect information from multiple sources, such as ERP, MES, HR, and purchasing systems. The AI acts as the intelligence layer between them, combining that information with the factory’s own planning rules to produce a recommendation or even drive production. Applied AI is the value we wanted to add: not just another report.
How does factory-wide AI work in practice?
MTN: One of the most interesting aspects of Epifor is that the platform can operate across an entire factory rather than being tied to one machine or process. How does that work in practice?
Giannis Savvaidis:
Epifor follows the order, not the machine. Every manufacturing order carries its own routing through the plant: laser cutting, bending, welding, assembly, quality, palletizing. At each work center, based on how the factory works, people report from a tablet or a smartphone with a couple of taps.
In labor intensive environments, knowing where the orders are is more valuable to manufacturers than machine data. This is also where we focus: labor intensive manufacturing.
What does factory-wide AI know that machine-level AI does not?
MTN: What does AI gain from having visibility across multiple machines, workstations and production processes that a machine-specific AI solution simply cannot provide?
Giannis Savvaidis:
Context. A machine level AI can tell you how a laser can be utilized at full speed with minimum energy. It cannot tell you that it is cutting parts for an order that ships next month, while a rush order waits in front of the press brake or the truck.
Our AI agents see orders, routings, priorities, delivery dates, capacity, downtime and quality together. With the whole picture, they answer the questions that drive decisions: which order is at risk and why, where the real bottleneck is and how many labor hours a work center needs next week to keep the schedule. In custom manufacturing with multiple steps, what matters most is that the factory delivers.
Can Epifor connect older machines and different ERP systems?
MTN: How does Epifor bring together data from different machines, PLCs, ERP systems and other shop-floor equipment, particularly in factories with equipment from multiple manufacturers and different generations?
Giannis Savvaidis:
We start with the production order, so the brand or age of a machine matters less than people expect. Orders come in from the ERP through our API, and progress, quantities and completions go back. It works in both directions.
Machines and PLCs can share data and send starts, stops, alarms and downtime automatically. However, we don’t do it very often. We gather the information directly from the operator or, if there is a SCADA system, we grab it from there. Different generations and different machine vendors are usually more of a headache than a useful insight in labor intensive manufacturing.
What happens when Epifor AI detects a production bottleneck?
MTN: Can you walk us through a real production scenario where Epifor’s AI identifies a problem or bottleneck across the factory and helps the production team respond?
Giannis Savvaidis:
Take a typical week in a sheet metal shop. Epifor’s forecasting flags that an order will finish after its delivery date. The production manager asks Epifor AI why.
The agents trace it across the floor: assembly needs 8.5 more hours on Friday than it has, because welding lost 7.5 hours earlier in the week to missing raw material, and the work stacked up behind it. Then they suggest options based on the plant’s own planning rules: which orders to group, what to run next and where to add hours, taking personnel availability into consideration too.
The manager decides and can ask the agents to update the schedule and change priorities. What used to require walking the floor and checking spreadsheets, becomes a short conversation.
How much autonomy does Epifor MES AI have?
MTN: Your AI agents cover areas such as production assistance, planning and reporting. How do these agents work together, and how much autonomy do they currently have when making or recommending production decisions?
Giannis Savvaidis:
Users ask one question in Epifor AI, which is equipped with the skills we provide. The AI grabs the user’s intent, and the right agent picks it up: production assistance for what is happening now, planning for what should happen next and reporting for how things went. They all work from the same live model of the plant and the same knowledge each customer adds, such as terminology, planning rules and work instructions, so the answers stay consistent.
On autonomy: we have worked countless hours, turning our knowledge from MES implementations in 350+ factories into robust skills to train our agents. So, they answer, they suggest, and when a user asks, they take planning actions such as changing a target date, a routing or a field in an order, within that user’s permissions. The decisions stay with the people who run production.
Does a manufacturer need to replace its existing ERP or machine monitoring?
MTN: Many manufacturers already have ERP, MES and machine monitoring systems in place. How difficult is it to introduce Epifor’s AI into an existing factory without replacing the technology and processes already being used?
Giannis Savvaidis:
Our AI is built into Epifor MES, so it arrives together with the execution layer. The ERP stays: Epifor connects to it in both directions and gives it a live shop floor. Where a plant already monitors its machines, that machine data can come into Epifor, which adds the order context around it.
The way the plant works stays too. We configure Epifor around the customer’s own routings, codes and terminology, and our consultants set up the agents around their planning rules. Go-live takes up to four months. Administrators need about an hour of training, and operators about five minutes.
People can also ask the agents how to use the system, in their own language, which speeds up adoption on the floor. And through our MCP connector, production data can reach the AI tools that teams already use, like ChatGPT, Claude and Copilot.
Epifor’s current website also describes two-way ERP integration and lists SAP, Odoo, Oracle and Infor among supported ERP environments, alongside API and MQTT machine connectivity.
Could AI eventually coordinate an entire factory?
MTN: Looking ahead, how far do you think factory-wide AI can go? Could we reach a point where AI continuously coordinates machines, people, scheduling and production decisions across an entire plant?
Giannis Savvaidis:
Yes, step by step. Our agents started by answering questions. Today they also suggest and, on request, take planning actions. Coordinating machines, people and schedules across the plant is the natural next step, but it needs work and a lot of testing from the factory side before it goes fully autonomous.
Three things will decide how far it goes: the quality of the data the factory puts in as knowledge, persistent testing and trust. That is why every Epifor project starts with real time visibility, and why people stay in charge of the decisions. At EuroBLECH we asked manufacturers which AI features they wanted next, and their answers are shaping our roadmap. AI will coordinate more and more of the plant, and the people who run production will make the calls.
Why it matters
The important part of Epifor’s approach is not simply that an MES now has a conversational AI interface.
It is the attempt to give AI enough production context to answer questions that cannot be solved by looking at one machine.
A late order may not be caused by the machine currently processing it. The underlying problem could be missing material earlier in the routing, insufficient labor at a downstream work center, an unexpected stoppage or a higher-priority job consuming capacity.
That is where factory-wide data becomes potentially valuable.
There is also a practical distinction for buyers. Epifor is not presented as a standalone AI layer that manufacturers simply add to any existing MES. Its AI is built into Epifor MES, meaning a company evaluating the technology is also evaluating the underlying execution platform.
That makes ERP integration, implementation effort, data quality, operator adoption, permissions and the suitability of Epifor for the manufacturer’s production model important parts of the buying decision.
MTN Analysis
Epifor’s strongest proposition is also what narrows its target market.
The company is concentrating on labor-intensive manufacturing, where tracking the progress of an order across multiple operations can be more important than extracting increasingly detailed information from an individual machine.
That makes the approach particularly interesting for custom and multi-stage manufacturing environments.
It also addresses a genuine industry problem. Rockwell Automation’s 2026 MES research found integration was both the leading buying requirement and a major modernization challenge, while 43% of respondents said they were not effectively using the data they already collect.
The important issue for prospective buyers is therefore not whether conversational AI can answer a production question in a demonstration. It is whether the underlying factory data is sufficiently complete, timely and reliable for that answer to support a real production decision.
Epifor itself acknowledges that limitation. Its AI documentation says answers are based on production information held within Epifor and that the system identifies situations where forecasts or other data are incomplete.
There are also commercial questions that remain unanswered by the interview. Epifor has not provided pricing here, so manufacturers would need to request a quotation based on their implementation. The interview provides an implementation period of up to four months, but buyers should verify what is included for their specific ERP, number of work centers, required integrations and AI configuration.
The bigger development to watch will be autonomy.
There is an important difference between AI identifying a scheduling problem, proposing a change and autonomously executing production decisions. Epifor currently keeps the user in that loop. Savvaidis says agents can make planning changes when requested and within the user’s permissions, while production decisions remain with the people running the factory.
For manufacturers assessing factory-wide AI today, that boundary may be just as important as the AI itself.
FAQ
What is Epifor AI?
Epifor AI is an agentic AI capability built into the Epifor MES platform. It uses production information and factory-specific knowledge to answer production questions, support planning and reporting, identify problems and, when requested by an authorized user, make certain planning changes.
Is Epifor AI designed to control individual CNC machines?
That is not its primary focus. Epifor follows production orders across work centers and is aimed particularly at labor-intensive manufacturing. Machine and PLC data can be integrated, but the platform’s emphasis is on factory-wide order and production context.
Does Epifor replace a manufacturer’s ERP?
No. According to Savvaidis, the existing ERP remains in place. Manufacturing orders can enter Epifor from the ERP and production progress, quantities and completion information can be sent back.
How long does Epifor take to implement?
Savvaidis says go-live can take up to four months. Epifor’s website says implementation can range from a few weeks to four months depending on digital maturity, with around 12 weeks quoted for a focused implementation. Buyers should obtain a project-specific implementation plan before purchasing.
Can Epifor AI make production decisions autonomously?
Not fully. The agents can answer questions, recommend actions and make certain planning changes when requested and permitted. Savvaidis says production decisions remain with the people running the factory.
How much does Epifor MES with AI cost?
Pricing was not provided in the interview and we have not found public standard pricing sufficient to establish what a specific manufacturing deployment would cost. Manufacturers should request a quotation covering the MES deployment, integrations, configuration, support and applicable AI capabilities.
Further Reading
For manufacturers evaluating Epifor MES AI, the official Epifor MES website provides further information on the platform, integrations and its approach to factory-wide production management.
For comparison with other approaches to connecting AI with manufacturing systems and production decisions:
Syspro Torque Brings Powerful AI Autonomy to Manufacturing looks at how Syspro is connecting AI agents with ERP, MES, SCADA and other operational systems.
LG CNS Factova: Connecting Factory Equipment and Putting AI to Work examines the practical challenge of connecting existing factory equipment and production systems before applying AI.
Bosch Explains How Agentic AI Could Change Manufacturing explores the wider move towards agentic AI and the question of how much autonomy manufacturers should give AI systems on the factory floor.
