Glossary

20 Agentic ERP terms every shop owner should know

Terms used for Agentic ERP and AI accounting assistants, each defined in a single sentence with a real accounting example. Full definition of Agentic ERP at /agentic-erp

Agentic ERP#: is an ERP system where AI agents take on back-office work, such as adjusting stock, receiving goods, balancing inventory across channels, summarizing tax, and preparing expense drafts, within approval levels that humans set.
ERP#: is a system that combines sales, inventory, purchasing, and accounting data in one database, so a confirmed order creates its accounting entries in the same database.
Back office#: is the internal work customers never see but the shop earns because of it, such as adjusting stock, receiving goods, recording expenses, and preparing PP30.
AI agent#: is a program that takes a job, plans, calls tools, and finishes it, such as answering this month's revenue versus last month plus receivables by calling several tools at once.
Agentic AI#: is AI that decides and acts within goals and tools set by humans, and refuses to guess when data is missing.
Nong Pao#: is Seller Pao's AI assistant in the bottom right of every screen, used to answer accounting questions and prepare document drafts for human review.
Chatbot#: is a program that talks to people in text based on a knowledge base, such as answering customers about delivery status.
Copilot#: is an AI assistant that drafts or suggests while a human always clicks, such as drafting a customer reply that a human sends.
LLM (Large Language Model)#: is an AI model trained on massive text that generates new text, such as parsing pay 3,500 baht for ads into a line name and an amount.
RPA#: is a technique where a program clicks and types instead of humans following a fixed script, such as downloading a file from the same button every morning.
Human-in-the-loop#: is a design where humans are the decision point at critical steps, such as every draft requiring human confirmation before saving.
BAL (Back-office Autonomy Levels)#: is Seller Pao's five-level framework measuring how much back-office work AI can do alone, from reporting (0) to full automation in a narrow scope (4).
Explainability#: is the ability of a system to explain why it answered or prepared something, such as showing the document-backed numbers behind an answer.
Hallucination#: is when AI generates plausible but false data; a well-designed system says it does not know instead of guessing.
MCP (Model Context Protocol)#: is a standard protocol letting AI call tools from many systems the same way.
WebMCP#: is a website exposing its tools over MCP for external AI, such as a user's agent pulling a tax summary straight from the system; see /webmcp
PDPA#: is Thailand's Personal Data Protection Act, defining how personal data may be collected, used, and shared lawfully.
PP30#: is the Thai VAT filing form that VAT-registered businesses submit every period; the system prepares drafts for review, but filing with the Revenue Department is human.
e-Tax Invoice#: is an electronic tax invoice submitted to Revenue Department standards.
General journal#: is the basic accounting record needing at least two balanced lines; the system prepares balanced drafts before a human posts them.