We have worked with a lot of companies over the years. One thing that we have found to be true time and time again is that companies don’t tend to discover data quality problems in NetSuite until it’s too late. They are mid-close, prepping for an audit, or trying to explain why a report looks wrong to a frustrated executive. By the time they find the issues the cleanup is painful and the trust in your data is shaken.
The problems usually aren’t dramatic. They’re small and quiet. Maybe test records that never got cleaned up — a few duplicate customers — some customers or vendors missing basic contact info. None of it breaks anything immediately — it just slowly strips away the reliability of all the downstream data and reporting inside your account.
As a part of our AI + NetSuite series, we ran a data quality sweep on a NetSuite account using Anthropic’s Claude.ai. Here’s what surfaced in under one minute.
Test Records in Production
The first thing we wanted to identify were customer names matching common test patterns — “test,” “demo,” “dummy,” “abc,” etc. When we asked Claude to search for these types of customers, we returned over 150 records. A significant portion of those were still marked active, meaning they show up in dropdowns, reports, and searches. Having a couple of test records in the system is totally normal — having 150 clogging up your dropdowns is a hazard.
In our experience, this isn’t unusual at all. It’s one of the most common things we find while doing health checks. NetSuite accounts accumulate test records over years of onboarding new hires and customizing your account. Nobody deletes them because nobody thinks to look.
You may question why we would use AI for this when you could simply export the data and find those patterns yourself. Of course you could do that. But you could also easily spend hours trying to sort through the test accounts named ABC Company and still incorrectly categorize the ABC Salvage Yard that’s a real customer. One thing we found was that Claude was able to go beyond the company names and look at all the data for common test data patterns to understand which ABC companies were real and which were not — and it did it in under one minute with a high degree of accuracy.
Duplicates
Next, Claude ran a query grouping active customers by company name and found over 260 names with more than one record. Some were obvious: placeholder names like “N/A,” “To Be Generated,” and “None.” Others were legitimate company names with two or three records, which means transaction history is split across them and your AR aging, reporting, and forecasting are affected accordingly.
NetSuite has its own duplicate detection, and it works pretty well for a lot of scenarios — specifically at the point of data entry. But we have found that AI is able to find patterns between records that you didn’t plan for when you set up your duplicate detection. These one-off duplicates can end up invading a system and causing havoc if left unchecked. This additional checkpoint can help make sure your data is rock solid going forward.
Missing Contact Fields
Another common issue with data integrity is active customers missing email, phone, or both. These are key data points for future sales, collections, relationship management, and more. To that end, we had Claude query the system for these scenarios and returned nearly 900 records. Some of these are expected — placeholder or anonymous customer records, for example (there was some crossover from our test data list). But the majority are real companies where someone just skipped the fields during data entry or migration.
Missing contact data isn’t just an inconvenience. This can dramatically affect downstream processes that rely on that information to function. Before long, you have automated collections processes failing silently and money going uncollected — all because of a missing email.
This is another area where you could, of course, run a saved search and push the data out to a spreadsheet for manual review. The power of this query is being able to use AI to easily prioritize the list based on recent sales or invoicing activity, open balances, and more — pulling together a list that lets you take action on your bad data rather than just stare at it.
Worth Running Quarterly
This kind of sweep shouldn’t be a one-time project. Test records build up over time. Duplicates find their way in through integrations or manual entry. Contact data goes stale. Running a quick AI-assisted audit quarterly — especially before period close — takes about one minute and catches problems that are going to come back and bite you when producing your period-end reports.
If you want to know what your account looks like, this is a good place to start.
Part of our AI + NetSuite series. Start with the pillar post: What Can AI Actually Do Inside Your NetSuite Account?
Want to know what other issues are hiding in your NetSuite account? We work with clients to run structured health checks that go deeper than this — record quality, configuration, role assignments, and more. Book a free consultation to talk through what makes sense for your setup.