We think we might be the first example of creating a NetSuite Purchase Order through an AI Conversation!

Here’s the setup:

  • NetSuite ERP
  • Salesforce
  • Breadwinner connecting Salesforce to NetSuite
  • Agentforce enabled in Salesforce

With that, we can have text interactions with the agent and ask them questions about data in NetSuite. So the Master Data system will be NetSuite, though thanks to Breadwinner’s replication technology, a live copy of this data is availabe in Salesforce.

Let’s start with a few questions about Inventory.

Keep in mind that for Inventory questions, we could expose this agent to internal emploees, channel partners, or even expose this on a customer-facing website.

Salesforce Agentforce Netsuite Inventory Questions

This is exciting! The agent can answer discrete questions such as the inventory in a specific location for a specific item, which is stored in a single field.

But the agent can also run live reports and aggregate data across all (or a subset of) locations. The agent can handle adding data, as well as min/max/ave, etc.

Even more impressive, the agent can have logic enabled to suggest inventory reorder points and ask about the creation of a purchase order.

Salesforce Agentforce Netsuite Purchase Order Creation

The agent goes ahead and, at the request of the user, and creates a Purchase Order in NetSuite. That PO is given the PO Number of “PO0000296” by NetSuite, and the Purchase Order record and its line items are replicated back to Salesforce for further action by either humans or agents.

Permissions and Security

Access to NetSuite data is handled by Salesforce’s Trust layer, which means the Agent will only provide data on records the Salesforce User could access based on their permissions.

And the ability to create Purchase Orders, or any other record in NetSuite, can be controlled through org-wide, per-user, or per-profile permission sets.

What This Means

With Agents, the way humans interact with Accounting, ERP and CRM will completely change. The old UI/UX is being eaten by agents. Guardrails and permissions will be held in the Agent, who will be able to handle business logic and spot anomalies.