AI for Epicor Prophet 21 (P21) — What If You Could Just Ask Your ERP?
Ask P21 Questions in Plain English. Explore the Data. Keep the Conversation Going.
What if getting information from Prophet 21 could start with a question instead of a report?
-
“Hey P21 — who are my top-performing customers this year?”
-
“Show me five years of sales for this item by month.”
-
“Based on those results, forecast the next six months.”
-
“Now factor in seasonality. What should I be buying right now?”
MindHARBOR is developing and pilot-testing a conversational AI solution that allows authorized users to interact with Epicor Prophet 21 ERP and related business data using natural language.
Instead of finding the right report, exporting data to Excel, waiting for someone to write a query, or starting a new analysis from scratch, users can simply begin with the business question.
The system can dynamically build the appropriate query, retrieve P21 data, visualize the results and provide the underlying information for review.
Then ask another question. The conversation becomes the analysis.
Example from MindHARBOR's current P21 AI pilot environment:
a natural-language question dynamically retrieves and visualizes underlying Prophet 21 data.
More Than an AI Chatbot
Putting a chat window in front of an ERP isn't particularly interesting by itself.
Giving users a new way to explore, understand and interact with their own ERP data is.
P21 ERP Chat is being designed to help users move naturally from a question to the underlying data—and then continue exploring what that information means.
A sales manager might ask who the top-performing customers are.
A salesperson could ask the same question and receive information appropriate to that user's authorized sales context.
A purchasing team might investigate item history, supplier performance or demand patterns.
Management could explore trends, exceptions and operational questions without knowing beforehand exactly which report, table or query contains the answer.
And the question doesn't have to end with the first result.
Ask. Explore. Drill deeper. Compare. Analyze. Ask again.
From “Show Me” to “Help Me Understand”
1. SHOW ME - “Show me five years of sales for this item by month.”
P21 ERP Chat dynamically queries the appropriate data and returns the results — including the underlying data and visualization.
2. ANALYZE IT - “Based on these results, forecast the next six months.”
Without starting over, the conversation moves into analysis — examining historical trends, seasonality, recent performance and assumptions.
3. HELP ME DECIDE - “With seasonality, what should I be buying right now?”
Now the conversation becomes a business decision. The AI evaluates what it already learned, identifies anomalies and provides recommendations along with the reasoning and caveats behind them.
Ask → Explore → Analyze → Decide
That's a very different experience from running another static report.
Your ERP Data Still Belongs to You
One of the first questions organizations should ask about enterprise AI isn't what model it uses.
It's: Where is our data going?
ERP systems contain some of a company's most sensitive operational information—customers, pricing, margins, purchasing, inventory, suppliers, transactions and business history.
MindHARBOR's P21 AI architecture is being developed with data control, privacy and security as core design considerations.
The solution can be deployed around a locally hosted AI server, allowing ERP data and AI processing to remain within customer-controlled infrastructure.
Where external AI services are appropriate, the architecture can also support third-party AI capabilities and controls around what information is provided to those services.
That gives organizations options rather than requiring their ERP data to be pushed indiscriminately into a public AI platform.
And because a local AI environment can support OpenAI-compatible interfaces, the same AI infrastructure may eventually support other approved business applications and AI use cases beyond P21 ERP Chat.
Your ERP data is your business. AI shouldn't require giving up control of it.
AI Without the Black Box
Getting an answer isn't enough.
Business users need to understand where the answer came from, what information was used, what assumptions were made and when human judgment is still required.
The current P21 ERP Chat environment is being developed to provide visibility into the information behind the conversation—including underlying result data, generated queries, analysis and assumptions.
The demo itself illustrates why that matters: when historical data contained conflicting seasonal signals, the analysis identified those anomalies and explicitly surfaced them as caveats rather than hiding them.
AI should help people make better decisions—not ask them to blindly trust an answer.
What Could You Ask P21?
Hey P21...
-
“Who are my top customers this quarter?”
-
“Which customers haven't ordered recently?”
-
“Show me sales for this product for the last five years.”
-
“Which items are declining compared with last year?”
-
“Who are my largest suppliers?”
-
“Show me unusual changes in purchasing.”
-
“Which products have the strongest seasonality?”
-
“Compare this year's sales with last year.”
-
“Which orders or customers should I be paying attention to?”
-
“Based on this history, what questions should I be asking?”
Those are examples, not predefined reports. The idea is to let the business question drive the interaction.
Built Around Your P21 Environment
Every Prophet 21 environment is different.
Companies have different data structures, customizations, business rules, integrations, security requirements, workflows and definitions of what matters.
That's why we're approaching P21 AI the same way we approach other MindHARBOR development and integration projects:
Understand the business problem. Understand the data. Control access. Build around the environment. Test the results. Expand where it creates value.
-
No giant AI transformation initiative required.
-
No assumption that every problem needs AI.
-
And no requirement to replace the P21 processes that already work.
Currently in Pilot Testing
We're kicking the tires ourselves—and inviting a limited number of P21 users to help us kick them harder.
MindHARBOR is currently developing and pilot-testing this capability with a small group of users as we continue refining the interface, data access, security, analysis and practical use cases.
If you'd like to explore what conversational access to your own Prophet 21 data could look like, we'd like to hear from you.
Looking for the Bigger AI Picture?
Learn more about Artificial Intelligence, Agentic AI, LLMs, RAG, MCP, data security and practical AI applications for Prophet 21 in our broader AI & Digital Intelligence resource center.
