What It Actually Is

The NetSuite AI Connector Service connects outside AI tools directly to your NetSuite account. Instead of Oracle building its own assistant, it built the plumbing — so you can bring the AI tool you already use, and it can pull data or take action inside NetSuite when you ask it to.

Nothing turns on by itself. Someone still has to set up the connection, decide which employees (and which AI tools) get access, and check what they can see.

It comes in three pieces:

Tools. The basic building blocks — the AI can look up records, run reports and saved searches, and pull or update data. It only sees and does what your NetSuite role already allows. An AI connected under a limited role can't suddenly see payroll or edit things it shouldn't.

A prompt library. Oracle also ships a library of over 100 ready-made prompts, organized by topic — things like AR aging, vendor fraud checks, working capital, fixed assets, and system health. You can copy them as-is or tweak them. It's meant to save you from writing prompts from scratch.

Skills. A shared set of NetSuite-specific instructions you can load into your AI tool so it understands NetSuite's terms and data structure better, instead of guessing.

Which AI Tool Should You Use?

Not all of them work equally well right now.

Claude is the safest bet today. Oracle names it directly as a supported tool, and even features a customer quote about using it:

"NetSuite AI Connector Service streamlines our data analysis and reporting processes with an easy prompt in Claude, enabling us to work faster and more efficiently."

— Michael Wollack, COO, Promier Products

In our own use, it's handled multi-step NetSuite questions cleanly and stayed inside its permissions well.

ChatGPT is also officially supported and works as a solid second option.

Microsoft Copilot isn't quite there yet. There's no simple, built-in way to connect it to NetSuite right now — it runs into a login/security step NetSuite requires that Copilot doesn't fully support out of the box. You can get it working with extra setup, but it takes real effort. If your team is Microsoft-first, budget time for that, or start with Claude or ChatGPT instead.

Where AI Actually Saves Time

Three uses come up again and again once a team starts using AI inside NetSuite.

Learning how to do something. If you don't know how to build a saved search, set up a workflow, or find a report, you can just ask. AI can walk you through it step by step, in plain language, instead of you digging through help articles or guessing.

"Walk me through, step by step, how to build a saved search that shows all open sales orders by customer, with a formula column for days past the promise date."

Questions that normally take three saved searches and Excel. Some questions — the crosstab-style ones, where you'd normally pull two or three saved searches and stitch them together by hand — AI can often just answer directly, in one shot.

Mass updates. This is where AI can save real hours: updating a batch of records at once instead of opening each one by hand. For example:

"For these 40 BOMs [list], remove the phantom assembly and list the raw components directly on the BOM instead."

"Update the component quantities on these BOMs to 120% of their current amount."

A word of caution on mass updates specifically: this is the one place worth slowing down. Ask the AI to show you what it's about to change before it changes anything, test on a handful of records first, and use a sandbox account if you have one. The same permission rules still apply — an AI connected under a read-only role can suggest the change, but it can't make it.

Prompts You Can Actually Test

These are examples to try once your connection is set up — not copied word-for-word from Oracle, just written in the same spirit for a manufacturer's data. Adjust field names to match your own account.

AR aging:

"Pull open AR aging by customer, bucketed 0-30/31-60/61-90/90+, and flag any customer whose 90+ balance grew more than 20% versus last month."

Vendor fraud check (the kind of thing behind our own fraud-detection work — see the case study on our homepage):

"Review vendor bills entered in the last 30 days for duplicate invoice numbers, near-duplicate amounts to the same vendor within 5 days, and any bill approved by the same person who created the vendor record."

Segregation of duties (see our article on this topic):

"List all active roles that can both edit vendor records and approve bills, and show which users hold each one."

Working capital:

"Calculate current DSO, DIO, and DPO for this month, compare to the trailing 3-month average, and explain the biggest change."

System cleanup:

"Find saved searches that haven't been run in 90 days, and flag any owned by users who are no longer active."

Inventory risk:

"Compare on-hand inventory against open sales orders and purchase orders for items below reorder point, and rank which ones are most likely to cause a shortage in the next 30 days."

One thing to know going in: if a prompt seems to "fail," it may just be hitting a permission limit, not a data problem — that's the system working as intended.

Permissions Matter More Than the Prompts

The most important part of this whole setup isn't the prompt library — it's that the AI only sees and does what the connected person's NetSuite role already allows. For manufacturers in ITAR, ISO, or other compliance-heavy environments, that's what makes this usable at all.

But "governed by roles" doesn't mean "safe by default." Someone still has to decide who gets AI access, whether it's read-only or read-write, and check it regularly — the same ownership question we've written about before with permissions in general.