# How to add an AI chatbot to Oracle NetSuite > Source: https://netray.co/integrations/netsuite-ai-chatbot > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: Integrations ## Short answer A NetSuite chatbot is a chat front end, either a Suitelet inside NetSuite or a page on your portal, that sends questions to a service which runs approved SuiteQL or RESTlet queries and answers only from the results. Role mapping decides who sees what, and the model can run on your own hardware. ## System Oracle NetSuite ## Capability ai-chatbot ## Intro Staff ask the same questions over and over: where is this order, do we have this item, what did we pay this vendor last quarter. Customers and suppliers ask versions of them by email. A chatbot grounded in NetSuite can absorb a large share of these, but only if it answers from live data rather than from a model's guess. NetSuite gives you two natural homes. Inside the account, a Suitelet can render a chat panel for employees. Outside, a customer or vendor portal, or a website that uses your own authentication, can host the chat while a backend service talks to NetSuite. The risk is not that the bot sounds wrong. It is that it sounds right while exposing a record to someone who should not see it. This page focuses on getting identity and grounding correct first. ## Use Cases ### Employee help desk for ERP questions **Description:** Staff ask how to create a vendor return or where a transaction type lives, and the bot answers from your internal procedures plus live record lookups. ### Customer order tracking **Description:** Customers on a portal ask about their orders and get status, tracking and invoice links limited to their own account. ### Supplier PO confirmations **Description:** Suppliers ask which POs are open and confirm dates in chat, which writes a pending note for the buyer to accept. ### Warehouse stock questions **Description:** Shop staff ask for bin quantities or the next receipt of an item from a tablet. ### Finance policy and approval status **Description:** Employees ask whether an expense report or bill is approved and who holds it. ## System Surface - Suitelets (SuiteScript 2.x) can host an in-account chat page and call your backend, with the session's NetSuite role available to the script. - SuiteQL over REST Web Services and RESTlets give the backend grounded lookups with explicit field lists. - NetSuite customer center and vendor center roles, or a separate portal with your own login, define external identity. - Custom records can hold chat transcripts, feedback and escalation tickets inside the account if you want them there. - SuiteFlow or User Event scripts can open a case or notify an owner when the bot hands off to a person. ## How It Works ### Pick the audience and surface **Detail:** Start with one group. Employees in a Suitelet is the lowest risk. External users need a portal with strong authentication. ### Map identity to data scope **Detail:** For employees, use their NetSuite role. For customers and vendors, filter every query by the authenticated entity ID in the backend, never by anything the model says. ### Build grounded retrieval **Detail:** Use fixed queries for live data and a document index for procedures and policies. Make the model cite the record number it used. ### Constrain the answer **Detail:** Instruct the bot to say it does not know when no source supports an answer, and to hand off to a person instead of guessing. ### Design the handoff **Detail:** Create a case or task in NetSuite with the transcript attached when the bot cannot help or the user asks for a person. ### Review transcripts weekly **Detail:** Read failed and thumbs-down conversations, then fix tools or documents rather than prompts alone. ## Example Flows ### Customer asks where an order is **Trigger:** A logged-in customer types, where is my order from last week? **Steps:** - The backend identifies the customer entity from the session. - A fixed query returns that customer's recent orders with fulfilment state. - The bot summarises the status and tracking number. - It offers a human handoff if the order is late. **Result:** The customer gets an answer instantly and customer service sees fewer status emails. ### Supplier confirms a date **Trigger:** A supplier in the vendor portal says they can ship line 3 on the 14th. **Steps:** - The bot looks up the supplier's open PO lines. - It confirms the line and echoes the proposed date back. - It writes a pending note on the PO. - The buyer receives a notification to accept or reject. **Result:** Confirmations are captured in NetSuite without a buyer retyping emails. ## On Prem NetSuite stays in Oracle's cloud, but the chatbot's model, retrieval index and transcripts can live in your data center or private cloud. The Suitelet or portal page calls your internal service, which calls NetSuite outbound and answers with an open-weight model on your GPUs. This matters when order content, part numbers or customer pricing are sensitive, or when your contracts forbid sending them to third-party AI vendors. For external users, put the chat service in a DMZ with web application firewall protection and keep the model itself in the internal zone. ## Security - Filter by authenticated entity ID in code. Never trust the model to scope customer data. - Exclude margin, cost and banking fields from every external-facing tool. - Add rate limiting and bot protection on public chat endpoints. - Store transcripts under a defined retention period and redact sensitive values. - Test prompt injection through free-text fields such as memo and comments, which are attacker-controlled in some flows. ## Options ### NetSuite's built-in assistive features **Fit:** Enough when users only need help text or simple record summaries inside the UI. **Effort:** Configuration only. ### Build with a Suitelet or portal and a backend service **Fit:** Good when you have SuiteScript and web developers and a single audience. **Effort:** One to three months including testing. ### Netray-built grounded chatbot with a private model **Fit:** Fits sensitive data, multiple audiences and a need for evaluation. **Effort:** A pilot with one audience, then add others. ## Pitfalls - Scoping customer data in the prompt instead of in the query. - Indexing documents that contain pricing for the wrong audience. - Letting memo fields flow into the prompt unfiltered, which opens prompt injection. - No handoff path, so users get stuck in a loop. - Launching to customers before staff have used it for a month. ## Faq ### Can the chatbot live inside the NetSuite UI? **Answer:** Yes, a Suitelet can render a chat panel for employees and pass their session to your backend. The exact placement depends on your UI and release, so prototype it. External users generally use a separate portal with your own authentication. ### How do we stop it showing the wrong customer's data? **Answer:** Enforce scope in code. Take the customer ID from the authenticated session, add it as a hard filter in every query, and exclude fields the customer should never see. Never rely on the prompt to apply that restriction. ### Can it run without a public AI service? **Answer:** Yes. Host an open-weight model and a retrieval index on your own infrastructure, and have the chat service call NetSuite outbound. Only data returned by the approved queries reaches the model, and transcripts stay with you. ### Can the bot create or change records? **Answer:** Keep the first version read-only. Later, allow narrow actions that create pending notes, cases or draft records, so a NetSuite user approves before anything affects transactions or inventory. ### How do we measure whether it works? **Answer:** Track resolution without handoff, thumbs ratings, wrong-record incidents and the drop in routine status emails. Review a sample of transcripts every week and fix the tool or document behind each miss. ## Search terms netsuite ai chatbot, chatbot for netsuite, netsuite chatbot integration, netsuite customer portal chatbot, netsuite suitelet chatbot, ask netsuite questions in plain english, netsuite virtual assistant, on premise ai chatbot netsuite, netsuite vendor portal chatbot --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json