Netsuite Agentic AI and MCP

This module explains what agentic AI is, how it differs from generative AI, and where it creates value in NetSuite implementations. You will learn the agent cycle; perceive, reason, plan, act, validate, and reflect along with single-agent and multi-agent designs, prompts, tool use, guardrails, and human-in-the-loop oversight. The module also covers NetSuite integration routes for agents, including SuiteTalk, SuiteQL, REST Web Services, and the NetSuite AI Connector Service (MCP), and helps you identify when agents are appropriate and when they are not.

Learn how the NetSuite AI Connector Service enables MCP-compatible AI clients, including AI assistants and agents, to work securely with NetSuite data. You will configure the required NetSuite features, including Server SuiteScript, OAuth 2.0, and REST Web Services; create a dedicated least-privilege agent role; and configure MCP Server Connection and OAuth token permissions. You will install and configure the MCP Standard Tools SuiteApp and Companion SuiteApp, connect an MCP-compatible AI client, and use natural language to query records, saved searches, reports, and SuiteQL data. The module also covers standard tool capabilities, limitations, and troubleshooting for connection, authorization, role, and tool-access issues.

Apply AI agents across the full NetSuite implementation lifecycle, from discovery through post-go-live support. You will use agents for requirements discovery and traceability, configuration gap-fit analysis, configuration reviews, data-migration validation, test-case generation, UAT evidence, defect logging, regression testing, and cutover readiness. You will work through agent use cases in Procure to Pay, including supplier onboarding, purchase orders, receiving, invoice-to-PO matching, exceptions, and approval routing; Order to Cash, including customer setup, sales orders, fulfilment, invoicing, credit checks, and collections; Record to Report, including journal entries, period close, reconciliations, intercompany transactions, variance analysis, and financial reporting; and Inventory, including item setup, reorder thresholds, stock counts, supply disruptions, and fulfilment exceptions. You will also learn how to define the boundary between agent assistance, human review, and formal business approval.

Build your own tools when the standard tools are not enough. You will set up an SDF SuiteApp or Account Customization Project (ACP) and write SuiteScript 2.1 Custom Tool scripts using the @NScriptType CustomTool entry point. You will define tools in JSON-RPC schemas, deploy and manage them through the toolset SDF object, and configure tool-level permissions and role access. The module covers input validation, error handling, logging, record and saved-search operations, SuiteQL, REST Web Services, and approval workflows. You will also build an MCP App, a self-contained HTML interface that renders inside the chat, and manage custom tools under Customization > Scripting > Custom Tools. Finally, you will learn version control, deployment, testing, documentation, and maintenance practices for custom tools.

Understand the risks of giving AI agents access to an ERP and apply the controls needed to operate them responsibly in production. Topics include least-privilege role design, separation of roles by business function and toolset, OAuth authorization and token management, role-based and tool-level permissions, approval workflows, sensitive-data controls, activity tracking, execution logs, audit trails, exception handling, and escalation rules for high-stakes or low-confidence decisions. You will also cover data privacy, vendor risk, AI-client due diligence, change control for prompts and tools, incident response, cost management, support ownership, monitoring, and ongoing maintenance.

Design, build, and demonstrate an agent-assisted process for one NetSuite ERP module or implementation scenario. You will select Procure to Pay, Order to Cash, Record to Report, or Inventory; define the business problem, target process, and agent use case; connect an MCP-compatible AI client through the AI Connector Service; use at least one standard MCP tool; and deploy at least one custom tool. Your capstone must demonstrate data validation, testing, approvals, exception handling, and audit controls. You will present the solution, governance model, risks, limitations, and implementation roadmap in a portfolio-ready format.