NetSuite

Configuration changes built-in settings, customisation extends NetSuite with supported tools, development writes code for what the platform can't do. The costly mistakes are treating a quick config change as a project, or a real build as a tweak. Here's how to tell them apart.

The wrong scope produces a confident, wrong number and native reporting won't warn you; on one board pack, out-of-scope entities inflated costs by ~£400k. Correct reporting curates and validates exactly which subsidiaries are in scope and builds that into the report.

Native search/list fields cap and hide valid matches, scheduled and Map/Reduce scripts hit governance limits and time out, and reports slow down, often invisibly. The fixes are targeted: purpose-built search tools, script optimisation and sometimes a data-model review.

The AI does only the extraction, turning a PDF into structured data; matching, GL coding, approval and the per-line audit trail all happen in NetSuite where finance governs them. It is human-in-the-loop, cost-and-control, not a black box.

Build automation that detects the parent-child hierarchy and produces one combined branded report instead of many. We built one consolidating 28 sub-accounts automatically, recognised by a report recipient on the parent, with no manual flag and no change to standard reports.

Build a system that generates each customer's report, applies your branding, and delivers it on schedule by email and portal. We built one producing multi-tab branded Excel workbooks plus a portal view, with automatic parent-account consolidation, for a supplier to UK local authorities.

Anchor it on accounting-line data, handle signs and item-level aggregation correctly, and reconciliation-test against P&L Detail before sign-off, all scoped with an FRD first. One we built this way reconciled to 0.12% on revenue.

Capture the emailed PO, extract the order data with the Claude API, and create the Sales Order in NetSuite via SuiteScript, handling customer-specific item codes, with review Suitelets and a feedback loop. Proven live against real customer POs.

Native list/record fields cap typeahead results, so at scale valid matches are silently hidden, causing missed or wrong quotes. A purpose-built Suitelet search returns every match; we built one across ~106,000 serialised assets with full history and a quote-to-order workflow.

Yes, without starting again. A fixed-price diagnostic finds and quantifies duplicates, orphaned records and broken linkage, then targeted corrective scripts fix them surgically with an audit trail, so you can trust your data again.