Every NetSuite account accumulates things that nobody owns. A workflow that was switched to not-initiating three years ago and never removed. A purchase order that's been open since a vendor went out of business. Six competing forms for the same record type, none of them marked preferred. A folder of integration dumps that's quietly pushing the file cabinet into overage fees.

None of it shows up on a report. NetSuite doesn't prompt anyone to look. So I've been building a set of prompts that do the looking, and they'll be in the next release of the Sonar AI Prompt Library. There are eleven of them, and I think of them as the account hygiene set. This post is a tour.

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 eleven are read-only. They recommend, and a person decides. That's deliberate, because disabling a deployment or deleting a file is destructive, and it needs impact confirmation that goes well beyond what an audit can see.

They also share a discipline about evidence. Each one distinguishes what it can prove from what it can only suspect, and it says which is which. A script with no log entries isn't proven unused, because scripts that log nothing by design leave no trace. A file with the same name and size as another is a strong duplicate hypothesis, not a confirmed duplicate, because you can't hash contents at scale. The prompts are written to make those distinctions rather than paper over them.

And most of them were built around a NetSuite quirk that had to be discovered the hard way. There's no form table exposed to SuiteQL. There's no reliable saved-search table either. Retention on execution logs is finite. Each prompt carries those verified facts so the agent doesn't burn turns rediscovering them.

Start With the Map

Account Configuration Summary is the document every new administrator, auditor, and implementation partner asks for on day one: a single, accurate atlas of how the account is configured. Enabled features, the subsidiary tree, the fiscal calendar, the user and role census, and which enabled capabilities are actually being used. It's built entirely from live introspection, and the rule is that it reports what it observed, never what accounts "typically" have. This one will be free.

NetSuite Utilization and Adoption Analyzer goes a step further and answers the question every NetSuite owner eventually asks: are we fully utilizing what we pay for? It distinguishes enabled from adopted. Features tell you what's switched on. Transaction and record volumes tell you what's used. The gap between the two is the finding, and every recommendation has to cite the number that justifies it.

Automation That Isn't

Zombie Automation Audit separates the automation that's configured from the automation that's alive. It cross-references every workflow and script deployment against runtime evidence and sorts them into tiers: structural zombies, which are configured not to run; contextual zombies, which are deployed against record types nothing touches; and silent ones, which are eligible to run but left no trace in the log window. Only the first tier is certain, and the prompt labels each tier so nobody mistakes "silent" for "unused."

Saved Search Ownership and Exposure Map treats saved searches as what they are: the account's shadow reporting layer. They feed dashboards, drive script logic, send scheduled emails, and encode business rules nobody wrote down. The prompt answers the governance questions. Who owns the load-bearing ones, what breaks when that person leaves, and which searches are single points of failure. A search counts as script-referenced only when its id appears in script source, which the prompt checks rather than assumes.

Form Customization Inventory catalogs the form layer, where customization sprawl hides: cloned variants nobody reconciled, bundle-installed forms shadowing in-house ones, record types with several competing forms and no preferred default. Since NetSuite exposes no form table to SuiteQL, the data path is extraction from the UI list pages, and the prompt only claims what those pages show.

Before Something Changes

Configuration Drift Snapshot and Diff builds the configuration changelog that NetSuite doesn't have. It captures a structured snapshot of the configuration surface, stores it as a timestamped JSON file in the file cabinet, and on every later run, diffs the live state against the last baseline. The snapshot deliberately captures stable identity, such as ids, names, flags, and statuses, and never volatile state like row counts. A diff full of self-changing noise is worthless.

Mass Update Pre-Flight Impact Analysis is the go/no-go briefing before anyone runs a bulk change. It pins down exactly which records are in scope, enumerates every user event script and workflow wired to that record type that a save could wake up, and grades how hard the change is to undo. It never performs the update, no matter how the request is phrased.

What Was Left Behind

Stale Open Transaction Groomer produces the groomed backlog: every open PO, sales order, and estimate that has outlived any plausible fulfillment horizon, aged and dollar-quantified, with a close-out recommendation per document. It's careful about partial processing. A PO with three of five lines received is open because of two lines, and the report shows line-level remainders rather than declaring the whole document stale. This one will be free too.

Required Field Compliance Audit finds the silent legacy of fields made mandatory after data already existed. NetSuite enforces a required field only on the next edit, so every record created before the change keeps its empty value. Those records break integrations, skew saved searches, and ambush users with forms that won't save. The prompt counts the violations per field, samples them, and proposes a backfill plan.

Deletion Forensics reads the one table that survives a record's deletion. System notes travel with the record, so when the record dies, its history goes quiet. The deleted-record table is the account's obituary column: type, name, who, when, and the mechanism. Most entries are routine. Severity comes from what was deleted, when, how many, and by whom, and the prompt is explicit that a deletion is an event, not an indictment.

File Cabinet Health and Growth is the storage dashboard NetSuite doesn't provide: size by folder and type, growth trend, the largest and stalest files, and duplicate-name clusters, ranked into a cleanup list that reclaims the most space at the least risk. Every candidate carries the caveat to verify usage before deleting, because a file's references aren't fully queryable.

Wrapping Up

None of these prompts will make a headline. They're the equivalent of walking the warehouse floor. But I've watched accounts carry years of this kind of debt because no one had a cheap way to see it, and the cost shows up later, in a release that breaks a zombie workflow or a bulk update that wakes up a script nobody remembered.

These eleven prompts will be in the next release of the Sonar AI Prompt Library. Account Configuration Summary and Stale Open Transaction Groomer will be free, and the rest will be in the paid tier. I'll post here when it's out.