A NetSuite account that's been running for a few years knows a great deal about itself that nobody has asked it. Which bundles have the biggest footprint. Which custom fields were created for a project that ended in 2021 and haven't held a value since. Where the file cabinet storage went. How many days a sales order actually spends becoming an invoice, and how many edits a transaction takes before it's left alone.

The Sonar edition of my NetSuite AI Prompt Library went live this month with a set of prompts in the Operational Intelligence category that ask those questions. This post covers the first thirteen, grouped by what they're asking about: the shape of the account, the quality of its data, and its rhythm.

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.

The Rule They Share

These are read-only audits. Every one of them ranks, flags, and recommends, and a person decides. A file is a deletion candidate, not a deletion. A field is unused until someone confirms it. That matters more for platform prompts than for finance prompts, because deleting a field or a file can break a script that nothing in the query layer can see, and the prompts carry that caveat on every candidate.

The Shape of the Account

Bundle and SuiteApp Inventory lists every installed bundle with its version, publisher, and footprint: the custom records, workflows, scripts, and fields it injected. Grouped by bundle, the footprint matrix shows which bundles are load-bearing and which are rationalization candidates, and it's the first thing to have in hand before an upgrade. Custom Record Type Sprawl catalogs every custom record type, attributes it to its bundle or to manual creation, estimates row counts for the largest, and surfaces the types that are empty or unused. The prompt handles failures gracefully, because some of those tables are empty or access-restricted, and it says which.

Custom Field Mortality is the audit I'd run first. It uses the custom field registry as the source of truth, separates bundle-managed fields from the ones people created by hand, and samples body fields one at a time to estimate fill rate, since there's no single query that measures every field at once. Fields that are unused, misnamed, or duplicating native fields come out as a prioritized rationalization list. Email Template and PDF Form Audit does the same for templates: every email template and advanced PDF or HTML form, with the dead weight, the broad-scope templates, and the formatting risk flagged.

Two more in this group were already in the library before the Sonar edition. Configuration Drift Snapshot and Diff builds the configuration changelog NetSuite doesn't have: a timestamped snapshot of features, roles, mandatory fields, workflows, deployments, and forms, stored in the file cabinet, and diffed against the last baseline on every run. It captures stable identity rather than volatile state, so the diff isn't full of self-changing noise. File Cabinet Health and Growth is the storage dashboard: size by folder and type, growth trend, the largest and stalest files, and duplicate-name clusters, with the reminder that a duplicate name is a hypothesis, not a proven duplicate file.

The Quality of the Data

GL Class and Department Tagging Quality measures how completely class and department are populated on GL lines, by account type, transaction type, and creator. The reason it matters is segment reporting: if the tags are blank, the segment P&L is unreliable, and the prompt labels any segment result built on a partial population until the tagging rate reaches 95%. It also carries a trap it found the hard way: class and department aren't readable from the accounting line table, so the join has to go through the transaction line.

Memo and Free-Text Anti-Patterns scans transaction memo fields for nulls, duplicates, numeric dumps, and compliance red flags, including plaintext payment details that should never be in a memo. It can stand up a recurring detection search as the remediation. Master Data Dormancy surfaces customers, vendors, and items that are still marked active but have had no transactions in twelve months or more, which is the list behind both a cleanup and a re-engagement campaign.

The Rhythm of the Business

Calendar Heatmap of Activity renders a year of transaction volume as a calendar grid, each cell a day, colored by intensity, weeks as columns and months labeled. Seasonality, dead zones, and crunch periods are visible at a glance in a way that a monthly bar chart can't show. Document Cycle Latency Atlas measures how long every document flow takes: sales order to invoice, purchase order to receipt, bill to payment, with the average and a P95 approximation for each, and bottleneck thresholds defined in the prompt rather than invented by the model.

Hidden Rework Tax mines the system notes audit trail for post-submission edit churn by transaction type. Every edit after submission is work that was done twice, and the pattern points at training gaps, weak validation, and unstable processes. Capacity Utilization Monitor measures throughput, production order volume, and departmental activity to find over- and under-utilized capacity. It's careful about which columns SuiteQL exposes for quantities and amounts, and it validates financial throughput against a native report before publishing.

Where to Start

If I were handed an unfamiliar account and an afternoon, I'd run them in this order. Bundle and SuiteApp Inventory first, because it tells you what you didn't build and can't easily change. Custom Field Mortality second, because dead fields are the cheapest debt to retire and the most common. Then GL Class and Department Tagging Quality, because if the tags are incomplete, every segment report the business relies on is quietly wrong, and that's a finding the controller wants today rather than at year end.

The rhythm prompts come last, and they're the ones to keep running. A calendar heatmap from one year is interesting. Two years side by side tell you whether the crunch periods are seasonal or self-inflicted. The latency atlas and the rework tax behave the same way: the first run establishes the baseline, and the value is in the runs after it.

Wrapping Up

None of these prompts will produce a number that goes in front of a board. What they produce is the list of things nobody owns, ranked by what it would cost to keep ignoring them. I've watched accounts carry that list unread for years, because the cost of finding it was an afternoon of someone's time. It's now a few minutes, which changes who's willing to look.

All thirteen are in the paid tier of the Sonar AI Prompt Library, under Operational Intelligence. Part two covers the next twelve.