Your ERP Already Has the Answers. AI Just Lets Your Team Ask.

What we showed at TUG CONNECTS 2026, and why bringing your own AI to your ERP beats waiting for a roadmap.

We put a question to the room at TUG CONNECTS this year:

How many of you have a colleague who avoids opening CSD? It’s just faster to call the person who knows where everything is.

Hands went up across the room. That reaction is the whole problem. The answers sit inside your ERP, and getting to them is the hard part.

The data is there. The access isn’t.

Infor CSD holds the data that runs your business: every customer, order, and release, inventory in each branch, the exact contract price you owe each account. Data was never the gap. Three barriers sit between it and your team.

ERP screens reward training and tribal knowledge. New hires take weeks to get up to speed, and the people who know where things live turn into bottlenecks.

A report tells you what happened. Someone still has to read it, compare it to last year, and decide what to do.

AI features ship every month, and ERP roadmaps move slower. They should. Your system of record needs to be stable, and chasing AI’s pace is somebody else’s job.

You don’t have to wait for your ERP vendor to add AI. You can bring your own, today.

Bring your own AI

An open standard called MCP, the Model Context Protocol, makes this possible. Think of it as a universal adapter: one protocol connects many AI tools to many business systems, so no one hand-builds a custom integration for every pair.

Two things matter here for a distributor. You choose the AI. Use Claude today, switch to another model next year, and your ERP vendor’s release schedule no longer sets your pace. And you govern it. You decide what the AI can see and what, if anything, it can do.

MCP is the protocol. Sirius is the bridge. Sirius is our ERP-aware layer, and it speaks CSD’s real data model: customer pricing, multi-warehouse inventory, release-level orders, ACL roles. Your AI talks to your ERP through one thin, governed layer. No screen-scraping. No custom API for every new question. CSD stays right where it is, as your single source of truth.

Three roles. One pattern.

Customer success: “Where’s my order?”

A key account calls about an order placed months ago. The promise date passed and the fittings never arrived. The CSR skips CSD and opens Claude instead, then works the problem out loud: pull the customer profile, find the stalled order, open the detail across releases, check invoices, check inventory across branches.

The CSR found two problems instead of one. A set of ball valves sat backordered sixteen months past the promise date, and a copper-fitting shipment went out in November that no one ever invoiced. No one was looking for that second one. The CSR opened zero ERP tabs during the whole investigation.

Sales management: “How is the month tracking?”

It’s Monday morning. Getting that read used to mean pinging an analyst, waiting for a dashboard, exporting to Excel, and building the pivots yourself. Three or four hours later, you’re left to decide what to do, alone. This time the manager types the question. Claude pulls five live ERP datasets in one conversation: top customers, revenue, year-over-year, pipeline, stalled deals, invoiced activity. It charts them on demand and sums up what to focus on this week. The whole thing took about fifteen minutes, and the manager spent that time deciding instead of assembling.

Inside sales: “Are these in stock, and what’s my price?”

The phone rings. A customer asks about two SKUs. The rep stays in one conversation: account health first, then live stock across thirty-one branches with bin locations and exact quantities, then the right price for that account, list price on one SKU and a 14% discount on the other, no rate-card lookup needed. The AI spots a related SKU the account keeps buying and flags it as an upsell. The customer had a confirmed quote in under three minutes.

A demo is easy. Production is the point.

It’s fair to be skeptical — plenty of AI demos fall apart the moment real data and real permissions show up. So here’s what makes this more than a party trick.

It’s read-only by default and governed by design. The permission structure is configurable. The AI sees what it’s allowed to see and nothing more, mapped to the same ACL roles you already manage in CSD.

It speaks your ERP, not just any API. The reason customer-specific pricing and release-level detail come back correctly is that Sirius understands those concepts natively. Generic connectors don’t.

And it’s the same bridge that already runs your storefront. Sirius wasn’t built for AI alone — the same integration layer drives live ERP data into Shopify, Shopware, and Adobe Commerce storefronts: real pricing, real inventory, self-service order and invoice history, even online payment of offline invoices. One investment, multiple surfaces.

Three things to take with you

Your ERP has the answers. Access was the problem, and MCP plus Sirius solves it.

You bring your own AI, so no vendor’s roadmap becomes your ceiling, including your ERP vendor’s.

Sirius makes it production-ready: governed, ACL-aware, and built on the same bridge that powers your storefront.

Our early-access program is live. If you’d like to watch your own CSD data answer questions like these, email hello@atwix.com and we’ll set up a walkthrough.