Some questions come around on a schedule. Is that blanket purchase order nearly used up? Why does this vendor bill not match the receipt? Which vendors are going to need a 1099 in January? Are the closed periods actually locked, or just closed? Every one of those has an answer in NetSuite, and every one of them usually gets answered by somebody exporting a search and doing the arithmetic in a spreadsheet.

When the Sonar edition of the prompt library went live this month, a good share of the new prompts went after exactly those questions. This post covers nine of them: three for purchasing after the order is cut, four for tax before the filing, and two that produce evidence an auditor can use.

If you're new to this, Sonar AI is an AI agent that runs inside NetSuite. Every prompt in the library is a playbook that I engineered and tested against live NetSuite data, and you run it inside your own account, against your own records.

What They Have in Common

All nine are read-only. They find, score, and recommend, and a person decides. Each one starts by resolving the account's configuration before it queries anything: how many subsidiaries, how many currencies, which status codes are in use. And each carries a short list of things that must go to a person. A vendor over the 1099 threshold, a price variance above 5% of the order, a period that should be locked and isn't. Those are flagged, not decided.

They also share a rule I've come to think of as the most important line in any of these prompts. Every figure traces to a query that was actually run. No vendor name, no payment amount, no coverage rate gets into the report unless the query that produced it is in the appendix.

Purchase Orders After They're Cut

Blanket PO Utilization Monitor answers a question that NetSuite's PO list doesn't. For every open order, how much of it has been consumed, and when does it expire? The prompt links vendor bills back to their orders, computes consumed against authorized, and sorts the results into over-utilized (billed more than the order value), under-utilized, and expiring. Over-utilization is a flag for a person, because it usually means either a bill was posted against the wrong order or somebody kept buying after the money ran out.

PO Receipt Variance is three-way match as an analysis rather than a workflow. It compares what was ordered, what was received, and what was billed, line by line, and it applies thresholds that are stated up front: a price variance above 5% of the order, a bill higher than the order, a receipt more than 30 days after the order, a bill with no receipt after 60 days. Each one has an action attached, and each action is a recommendation for the buyer or the receiving team, not something the prompt does itself.

Vendor Performance Scorecard gives every vendor a composite score from 0 to 100 across four dimensions: spend, PO volume, bill discipline (how long after the order the bill arrives), and concentration risk. The weights are in the prompt, and so are the tier cutoffs. A vendor above 25% of total spend scores low on concentration no matter how good the rest looks, because that's the vendor you can't afford to lose.

Tax, Before the Filing

Sales Tax Trend by State reads collected and accrued sales tax by state over time, and it looks for three things: states where activity has crossed a nexus threshold without a registration, rate changes that show up in the data before anyone updated the configuration, and filing exposure by period. It joins invoice and cash sale lines to the tax configuration, and it's explicit that it can't see a registration number that NetSuite doesn't expose to a query, so it loads the record when it has to.

Tax Code Coverage Audit answers a simpler and more uncomfortable question: how many revenue lines have no tax code at all? A blank tax code means tax was never calculated, and the prompt measures the coverage rate and grades it. Above 95% is adequate. Between 85% and 94% means reviewing exemption certificates. Below 85% is a remediation plan, and the prompt marks that conclusion for confirmation with a tax advisor rather than treating itself as one.

Tax Filing Readiness is the pre-filing checklist. It confirms that sales tax payable balances match what was collected, that nexus coverage is complete, that the periods being filed are closed, and that the 1099 vendor list is complete. It can cross-check against NetSuite's own Sales Tax Liability report when a native view is the authority. The output is a readiness checklist with a pass or fail on each item and the query behind it.

1099 Preparation Assistant identifies 1099-eligible vendors, aggregates year-to-date payments to each, and flags the ones above the IRS threshold along with any compliance gap, such as an eligible vendor with no tax ID on file. It has a hard rule I want to call out. It never writes a taxpayer identification number into its output. It masks or omits them, because a report that gets emailed around shouldn't carry them.

Evidence for the Auditor

Internal Controls Compliance Review (SOX/SOC) runs a structured set of control tests against the COSO 2013 framework, with live evidence from the account. The first test is whether closed periods are locked, and it's a good example of how the prompt thinks. It doesn't ask whether the period is closed. It asks whether every past period is locked for all users, because a closed period that isn't locked can still take a posting. Each test records pass or fail with the query that produced it, and there's a rule that a query returning zero rows doesn't prove a control passed until the query logic itself has been confirmed. The deliverable is a controls assessment matrix.

Address Quality Audit is humbler, and I included it here because addresses are the master data nobody looks at until a shipment or a tax calculation goes wrong. The prompt computes null rates per field on customer and vendor addresses, lists every malformed ZIP code individually (the leading zero lost in a spreadsheet import is the classic), and flags populations where every record says "US" when some of them plainly aren't. It also knows the difference between a deliberately minimal address on a cash-and-carry account and a real gap.

Wrapping Up

None of these prompts does anything a good analyst couldn't do with a saved search and an afternoon. What they change is the cost of asking. When the answer to "is that blanket PO nearly used up" takes a minute instead of an afternoon, people ask it every month instead of once a year, and that's when the answer starts to matter.

All nine are in the paid tier of the Sonar AI Prompt Library, under Expense and Cost Management, Tax and Compliance, and Audit and Assurance.