I was showing NetSuite’s Lot Traceability tool (a free bundle, in case you’re interested) to a customer recently who asked “We’ve got an AI connected to our NetSuite using the MCP connection - couldn’t I just use that to do a trace for me?”
There are a couple of things to consider when you reach the “to use AI or not to use AI” decision.
The first is value. NetSuite Next’s Ask Oracle consumes NetSuite AI Units, with usage varying based on the processing required. Even small requests repeated throughout the day or across a team can add up, so it makes sense to consider where that usage delivers the most benefit.
If a familiar list, saved search or report already gives you exactly what you need, the better option is usually to use that tool rather than call on Ask Oracle and consume some of your AI Units. If you’re using an external MCP connection, you have to consider how your usage is priced.
In this case, Ask Oracle is a great tool, and will let you get the information you need without knowing how to create elaborate searches. But remember: asking the same question again may burn a few AI Units (it depends on exactly how you execute the re-query). As a result, whenever you use Ask Oracle to find records, consider converting the results to a saved search. How would you do that? You can use Ask Oracle – once it shows you the results, ask it “can you make that into a saved search for me?” Not every Ask Oracle result can be converted to a saved search, but if it can, you use your AI Units to develop the search, then run the saved search directly for future inquiries, without repeating the Ask Oracle request.
Even if your AI Unit or token burn rate aren’t a concern, you should still consider consistency of results. For recurring work like lot tracing, I want established criteria that I can review and repeat. A conversational request can leave room for interpretation about which records, relationships, or date ranges to include. Things like AI skills help hedge against this phenomenon but isn’t a pre-existing Lot Traceability tool kind of the ultimate “lot trace skill” already?
It’s also worth noting the value of time-saved. I don’t want to discount that. If it’s quicker and easier to use AI, that’s should be part of your mental calculus.
So the short answer is to lean toward existing tools, especially if they’re doing precisely what you need them to do and particularly for repeatable, operational work. Use AI – whether Ask Oracle or an MCP Connection - when its ability to understand a question, bring information together, or help explain and interpret results adds value.
And use them in tandem when appropriate. For example, the Lot Traceability tool can produce the trace, while AI may help you summarize the findings.
Let us know if you've established any rules of thumb for when AI should and shouldn't be used.