This article was originally published on LinkedIn.

Over the past few months, several new "AI for NetSuite" products have been introduced into the market. They promise smarter automation, cleaner insights, and boardroom-ready analytics, all powered by proprietary systems that sit between you and your data.

They sound impressive. But do NetSuite customers really need another system to get meaningful value from AI?

I don't think they do. Not now, and maybe not ever.

The Power We Already Have

In August, NetSuite released something extraordinary: an official, secure bridge between NetSuite and the world's leading AI models.

The NetSuite AI Connector allows you to:

  • Make real NetSuite data available to models like ChatGPT or Claude - securely and on demand.
  • Receive structured, auditable insights in return.
  • Operate entirely within NetSuite's governance and security model - roles, permissions, logs, and controls.

The AI Connector is much more than a new feature. It's a platform-level capability on the same scale as RESTlets or SuiteTalk REST.

When the Connector launched, it effectively ended the need for SuiteAnalyzer, an external AI system I spent most of the summer building. The connector made it unnecessary.

And that's the point: we now have the right foundation. The only missing piece is how we use it.

The Temptation of Third-Party "AI for NetSuite" Platforms

Some of the third-party "AI for NetSuite" tools are genuinely innovative. Others are just repackaged chatbots. But almost all of them introduce technical layers that NetSuite customers don't really need.

And each new layer comes with hidden costs:

  • Data Exposure: Your financial data now lives on someone else's infrastructure.
  • Lock-In: You're tied to their roadmap, their pricing, their interpretation of NetSuite.
  • Lag: Their platform becomes the bottleneck as AI models evolve faster than they can adapt.
  • Cost: You effectively pay twice — the LLM fee, plus an additional charge for the intermediary platform that sits in the middle.

For most NetSuite customers the tradeoff simply isn't worth it.

The Hidden Risk: Your Data Leaves NetSuite

There's another problem with many of the "purpose-built" AI platforms, and it's a problem that often gets glossed over in the marketing.

To function, most of these systems must copy your NetSuite data into their own infrastructure. That usually means:

  • Syncing your transactions into a vendor-controlled database
  • Storing historical data outside your system of record
  • Running your queries or analyses through their servers
  • Retaining sensitive information for "context" or "training"

Even when encrypted, this creates material risk:

  • New attack surfaces appear outside NetSuite's security perimeter.
  • Backups and shadow copies may persist long after you believe data has been deleted.
  • Vendor employees may have indirect access to infrastructure or logs.
  • Audit trails and compliance boundaries become more complicated, especially when it comes to SOX, HIPAA, GDPR, SOC2, and industry-specific regulations.

Finance teams have spent decades keeping their data inside NetSuite's governance model for a reason. Every time the data leaves, the risk profile expands.

With the NetSuite AI Connector, the data path is simple and secure:

NetSuite → Oracle-secured AI request → AI model (ephemeral) → Response

No replication. No third-party storage. No vendor database.

Your sensitive financial, operational, and CRM data remains exactly where it belongs: inside NetSuite.

Most purpose-built AI platforms can't offer that. It's not because they're negligent, but because their architecture requires your data to leave NetSuite. That's a risk most companies simply can't afford to take.

The Myth of "Purpose-Built, NetSuite-Trained AI"

Many of the new AI platforms insist that generic AI "doesn't understand NetSuite," and that only a "purpose-built, NetSuite-trained model" can produce accurate or contextual results.

That sounds convincing, but it's simply not true.

AI doesn't need to be pre-trained on NetSuite to work with NetSuite

Modern AI systems don't need to memorize NetSuite's schema to analyze NetSuite data. They only need the right data at the right time, which is what the AI Connector provides.

Once the model has the structured data:

  • It analyzes it.
  • It reasons over it.
  • It generates insight.

Training the model on NetSuite's entire data model is unnecessary.

Retrieval beats pre-training every time

The current best practice in AI architecture is:

Don't train the model on the system. Retrieve the data the model needs from the system.

This is exactly what the AI Connector provides. A clean, real-time retrieval pattern aligned with how modern LLMs are designed to operate.

"Purpose-built" often just means pre-written prompts and UI polish

When you look closely, most of these platforms aren't really running custom-trained models. Instead, they're actually running the same LLMs you already have access to, and they're simply wrapped inside a proprietary UI or workflow engine.

The "NetSuite-trained" claim is often marketing language for:

  • A library of SuiteQL queries.
  • A set of prewritten prompts.
  • Some hard-coded domain logic.
  • A middleware layer.

You can reproduce and improve upon those things yourself by using the Connector and disciplined prompt engineering.

Customization makes pre-training brittle anyway

No two NetSuite accounts look alike. Segments, scripts, custom records, workflows, industry-specific setups. The list of potential differences between instances goes on and on.

A "pre-trained NetSuite model" can't possibly predict all of those variations. But the Connector's retrieval model avoids this issue, because it pulls data directly from your environment in real-time.

Why Building Your Own AI Layer Isn't the Answer Either

Some companies are experimenting with building their own AI execution layers, including MCP servers, new SuiteScript "Custom Tool" integrations, and other homegrown systems.

These experiments are interesting, but:

  • Most finance teams don't have AI engineering talent
  • Most companies can't maintain custom AI infrastructure
  • And the speed of AI progress will outdate custom builds before they stabilize

As the saying goes, "the juice rarely justifies the squeeze."

Work With What We've Got

Since the release of the NetSuite AI Connector, my approach to NetSuite AI has been simple:

Don't build around NetSuite. Don't build on top of NetSuite. Build with NetSuite.

For CFO-level analysis, the formula is already available:

  • The AI Connector for governed, secure data access
  • A leading LLM for reasoning
  • A library of high-quality prompts for structure, accuracy, and repeatability

That's it.

No lock-in. No middleware vendor. No proprietary "NetSuite-trained" claims. Just clean, transparent AI prompts that anyone can run.

Wrapping Up

NetSuite has already given us a secure, future-proof foundation for AI. And that foundation gives you something far more valuable than any third-party platform:

Freedom.

Freedom to evolve with the models. Freedom to design your own workflows. Freedom to avoid lock-in and infrastructure risks you don't need.

So before you invest in an AI-for-NetSuite platform, ask yourself:

  • Does this give me capabilities I cannot already achieve through the AI Connector?
  • Or am I outsourcing something that prompt design and disciplined retrieval already solve?

NetSuite's AI advantage won't hinge on external platforms. It will be realized by organizations that fully leverage the native tools the platform already delivers.