ERP Vertical AI Agents Aim to Simplify Access to Answers for Quick Action
Epicor Software Corporation has launched Epicor, a network of AI-driven vertical agents designed for supply chain industries. Integrated with Epicor Kinetic, a platform for manufacturing, in the Epicor Industry ERP Cloud, Prism enhances conversational ERP, allowing businesses to quickly access insights and recommendations through natural language queries.
Built on the company’s industry-specific data structure, Prism is made to streamline processes such as automating supplier communications, generating business process code and accessing logistics insights. Early adopters, like Madsen’s Custom Cabinets, anticipate time savings and improved efficiency by automating routine tasks.
READ MORE: Epicor Introduces AI and BI Capabilities with its Grow Portfolio
“For example, we spend many hours each week manually updating hundreds of line-item due dates that correlate to the jobs we have. With Epicor Prism, we believe we’re going to free significant time for our team to focus less on scheduling and more on production and meeting customer demand. That’s really adding value to our business,” said Tyler Madsen, director at Madsen’s Custom Cabinets, a Canadian manufacturer, in a press release.
To learn more about what Epicor Prism means for the Machine Design audience, we reached out to Michael Atkisson, Ph.D., director of product management, AI, at Epicor.
Editor’s note: Questions and answers may be edited for style and space.
Machine Design: How do vertical AI agents differ from general AI systems in the context of manufacturing ERP?
Michael Atkisson: Vertical refers to the idea that agents are specialized for the industry you operate in—e.g., manufacturing, distribution and retail; Prism agents are built for these specific industries. Agents we have built for the Kinetic manufacturing ecosystem differ from those we are building for our Prophet 21 distribution platform.
When it comes to generic generative AI like using ChatGPT and Microsoft Copilot, they don't have any awareness of your manufacturing ERP system data or your other private and proprietary data. They only know what you submit to them in the chatbot window.
With Epicor Prism, on the other hand, Kinetic’s 30-plus years of experience in creating the business backbone of manufacturing companies is built into the application. Prism has the context not only of Epicor's standard data structures but also the customizations you have made to the system.
Additionally, what you submit to Epicor Prism won’t train OpenAI's next version of ChatGPT with your proprietary data, which is what happens when you submit your data to personal ChatGPT accounts.
MD: Can you explain how these vertical AI agents are being integrated into existing manufacturing ERP systems?
MA: We launched Epicor Prism in January of this year. It has various agents to answer help questions on how to carry out tasks in Kinetic, you can ask it questions about graphs, charts and images of parts and products. You can save time by asking it for data from the system in your own words.
We’ll soon be launching an agent for ECM, our content management system, that will enable you to retrieve data from your stored documents without having to go to the CMS—you will just have to ask Prism.
In our next phase of development, we will be introducing automations to enhance time savings through on-demand bulk actions and by adding Prism to our ERP development tools. Customers will be able to use Prism to create dashboards, saved queries and the like.
READ MORE: Looking Back at IMTS 2024: A Vision for Manufacturing Innovation in 2025
MD: In what ways can vertical AI agents contribute to optimizing design processes for engineers? Can you provide specific examples?
MA: One way Prism can help today to get information to engineers faster is its ability to understand the information you are asking for, saving you from having to go to various places to retrieve it. We are exploring other capabilities directly with our customers and building agents to solve specific challenges.
For example, one customer asked for an agent that would search semantically on their product images and specifications in Kinetic to accelerate showing similar examples to what their customers had asked for. We are beginning the development of that use case with them. This will shorten the lead time of build-to-order scenarios, make faster quotes and ship product more quickly.
MD: How do these AI agents facilitate data analysis and decision making for design engineers?
MA: Prism today can analyze your dashboards, tables and other data representations. It can also pull information quickly from the system because of its knowledge of our data structure. Without Prism, you would have to understand how to write in a query language or click through a considerable number of filters. This saves design engineers’ time. We will be adding more analysis capabilities and triggered actions as we develop Prism.
MD: How do vertical AI agents facilitate a feedback loop from manufacturing outcomes back to design engineers for continuous improvement?
MA: We are working with customers to make a series of agents that auto-tune a variety of systems in Kinetic, like MRP. For example, the historical data of what has happened in the supply chain and facility will inform the recommended inputs to MRP. When what’s happening on the ground in real life gets added to the parameters of the MRP with ease, the plans will consistently be much more accurate and useful, driving important metrics such as inventory turns.
