Acumatica's AI features are real. Now what?
Acumatica 2026 R1 shipped AI Studio as generally available. AI Assistant is in experimental access. Most customers upgraded and enabled the features. Very few configured them correctly.
We help you decide what's worth turning on, set it up so it actually delivers, and build custom agents where native features stop short — including connecting Claude directly to your live tenant so your team can create orders, pull reports, and manage operations in plain English.
What we actually do with AI for Acumatica customers.
Six things, in roughly the order they matter.
01
Decide what's worth turning on
Most companies don't need every AI feature. We help you pick the highest-leverage ones for your business and ignore the rest.
02
Connect Claude directly to your Acumatica tenant
Using the CData MCP Server, we connect Claude to your live system. Your team types a sentence and gets an answer from real data — or creates a record. "Create a sales order for customer ABC, 50 units, ship by Friday." Done. No clicks. No ERP training required. Read and write, against your live tenant, with guardrails in place.
03
Build custom agents where native isn't enough
Order creation, invoice processing, 3PL reconciliation, cross-system workflows. Where AI Studio and AI Assistant stop, we pick up with Claude and MCP. We've built these in production. Same architecture handles sales orders and POs.
04
Configure AI Studio and AI Assistant well
The defaults work. The prompts and Generic Inquiry descriptions don't tune themselves. We do that work so the features deliver real value.
05
Set up evals and guardrails before production
Nothing that writes to your system ships without monitoring, review checkpoints, and a way to roll back. We move slowly on purpose.
06
Ongoing operations
New model releases, new prompts, new workflow candidates. Managed on retainer or per sprint.
