NetSuite expands toolkit to ease enterprise use of third-party AI assistants with ERP data

NetSuite is expanding its AI Connector Service with what it calls Companion capabilities its customers can use to hook up AI assistants with their choice of ERP data.

The update introduces prebuilt prompts, role-based controls, and domain-specific “skills” that help external AI systems better understand NetSuite’s data structures, workflows, and permissions with the help of Model Context Protocol (MCP). Under the hood, MCP exposes NetSuite data and actions in a structured format, allowing the Companion components to broker interactions between the ERP system and external AI assistants, the company said, the company said.

Companion capabilities aim to help scale AI pilots

The additions could help address some of the practical hurdles that enterprise teams encounter when trying to move AI initiatives beyond early-stage experimentation, analysts said.

“When a CFO connects Claude or ChatGPT to their NetSuite instance, the AI model doesn’t inherently understand what a ‘subsidiary consolidation’ means in NetSuite’s data schema, which fields map to which roles, or what governance rules apply to a Treasury Analyst versus a Controller,” said Ashish Chaturvedi, leader of executive research at HFS Research.

The Companion capabilities fill that gap, which is where many pilot projects stall; by standardizing prompts and embedding domain context, NetSuite is trying to make interactions with third-party AI assistants more predictable and enterprise-ready, he said.

The new Companion capabilities also help engineering and deployment teams eliminate a significant chunk of the “translation layer” work that integration teams would otherwise have to build from scratch when integrating third-party AI assistants with NetSuite ERP data, according to Robert Kramer, principal analyst at Moor Insights and Strategy.

That, Kramer said, could help CIOs accelerate early deployments while improving consistency and reducing the risk of shadow AI emerging across the enterprise.

However, Scott Bickley, advisory fellow at Info-Tech Research Group, noted that the new capabilities are unlikely to eliminate the need for orchestration work entirely: “Technical teams will still need to enable SuiteScript and OAuth, assign permissions, connect to the AI clients along with other integration management tasks related to token consumption limits and monitoring, log management, and any customization required to meet bespoke company workflows.”

Flexibility comes with trade-offs

Even with these limitations, Chaturvedi sees NetSuite’s strategic positioning of the new features transcending their individual benefits for CIOs.

“NetSuite is making a deliberate bet that the AI layer should not live inside the ERP. It should live wherever the customer wants it, with the ERP serving as a governed data source that any AI assistant can connect to,” Chaturvedi said.

But he cautioned CIOs about the governance model of NetSuite’s “bring your own AI” business model.

“The approach inherently distributes risk,” he said. “When an AI assistant connected through MCP generates a bad financial recommendation or surfaces incorrect data, the accountability chain is murky. Is it NetSuite’s fault for how the data was exposed? The AI provider’s fault for how the data was interpreted? The user’s fault for how the prompt was framed?”

NetSuite is not alone in taking this approach, and other vendors are explore it too: Microsoft offers a variant of NetSuite’s new capabilities in Dynamics 365 ERP, and Salesforce offers “ready-made” MCP servers for Salesforce, Heroku, and Mulesoft.

SAP and Workday, too, deploy MCP connections via their Joule Agent Hub and Agent Gateway respectively.

Where NetSuite differs, said Info-Tech Research Group’s Bickley, is that “It combines several pieces into a single finance oriented package — ERP connectivity, prompt library/scaffolding, UI widgets — making this connector service likely easier to deploy and potentially more useful and maintainable.”

Filling the prompt gap

As part of the updates, NetSuite also introduced a set of MCP-based applications.

Called NetSuite MCP Apps, these are essentially structured UI components such as forms, selectors, filters, and menus that render inside an external AI assistant such as Claude, ChatGPT, or Gemini but look and feel like native NetSuite interfaces.

Enterprises can use these interface elements to overcome what HFS Research’s Chaturvedi calls the “prompt gap,” caused by business users not knowing what to ask, how to ask it, or what data structures could provide answer. “MCP Apps close that gap by replacing freeform prompting with guided, point-and-click interactions.”

For Bickley, MCP Apps’ ability to merge external LLMs with NetSuite data can provide guided workflows that are more usable for non-experts, increasing internal adoption of AI in an enterprise, in turn helping scale pilots to production.

Extending AI beyond transactional ERP data

Other updates that NetSuite introduced include the NetSuite AI Connector Service for NetSuite Analytics Warehouse, which it said will help enterprises extend AI access beyond transactional ERP data to incorporate historical, analytical, and third-party data.

For enterprises, the practical effect is that AI can now bridge the operational and analytical layers without requiring a separate data pipeline or a different tool, said Chaturvedi.

But, cautioned Moor Insights and Strategy’s Kramer, keep an eye on data consistency: “If definitions don’t line up across systems, AI just scales the confusion faster.”

The Companion capabilities and the AI Connector Service for NetSuite Analytics Warehouse are now available globally in English, with additional language support planned.

MCP-based applications, NetSuite said, will be released as part of the MCP Standard Tools SuiteApp and made available through the SuiteApp Marketplace.