Happy Monday, everybody — Albert here, back in a proper office for the first time in about five years, broadcasting from a new space right in downtown Traverse City; we’ve got two big tech stories worth your attention this week — Sage 100 just shipped an on-premises AI module for predictive maintenance, and Microsoft Copilot can now chat directly with your Epicor Kinetic ERP data. Let’s get into it.
Or watch on YouTube.
Sage 100 Ships On-Prem AI for Predictive Maintenance — and That “On-Prem” Part Actually Matters
Sage released a new AI module for Sage 100 aimed at manufacturers. The pitch: it predicts equipment failures before they happen by integrating with your existing shop floor sensors and running extrapolations on the data to figure out when something is going to go sideways, not just that it already did. The module runs entirely on local hardware. Your data never leaves your servers. No vendor cloud involved.
That last part is the part I want you to actually notice.
We spend a lot of time talking about AI in this newsletter, and the conversation almost always assumes the cloud. But there’s a real, legitimate reason to want your data processed on your own iron — and it’s not just paranoia. If you are feeding a public AI your proposals, your pricing, your job costing, your customer lists, you are training that model. The model doesn’t put your name on the output when your competitor asks a similar question. But your pricing logic? Your margin structure? That’s in there. It’s informing the answers. You may never see it happen, but it’s happening.
At a minimum, you should be on an AI subscription that explicitly does not train on your data. In a lot of cases — especially if you’re a manufacturer with sensitive quoting or process data — you should be running AI locally. We’ve got our own internal AI box here at ACS (it’s currently in a moving truck headed north, but that’s a different story). The point is, this is solvable. It doesn’t require a Sage 100 module to solve it, either.
Which brings me to the part of this announcement I find genuinely interesting from an engineering standpoint: the fact that a major ERP vendor is shipping an on-prem AI module at all signals something. For years, every software company’s answer to everything was “put it in our cloud.” Seeing Sage go the other direction — local model, local processing, local data — is a real acknowledgment that their manufacturing customers have real security concerns and aren’t going to just wave them away because the UI looks nice.
Credit where credit’s due: it took them a while to get here. AI has been a thing for several years now. The first wave of integrations was basically just a chat window bolted onto the side of the application. This is more interesting than that. Predictive maintenance from sensor data, processed locally, surfaced inside the ERP you’re already running — that’s a workflow that could actually change how a shop floor operates. We’ll see how it performs in practice. But the architecture is right.
The question I’m watching: homegrown AI setups versus developer-baked modules like this one — which actually wins in the field? My vendors are all racing to embed AI into their tools. I’ve already built a lot of those integrations myself. It’s going to be an interesting race.
Copilot Can Now Chat With Your Epicor Kinetic ERP — and This Is What “AI in Your Software” Should Actually Look Like
The second story comes from ZDNet. As of June 25th, Microsoft Copilot can query Epicor Kinetic ERP directly. You can ask it questions about your data in plain language and get answers back — with context you might not have even thought to ask for.
Alert readers will recognize why this is a bigger deal than it sounds.
Getting data out of ERP systems has always been the unglamorous problem nobody talks about in the sales demo. The data is in there. But to get a useful answer out, you historically needed to know the reporting module, or write a query, or hire someone who knew the reporting module, or build a custom application around the reporting module — and even then you were constrained by what the report designer thought you might want to know. The gate was high. Most users just didn’t bother and made decisions on gut feel instead.
What Copilot integration does is lower that gate to basically zero. You type a question. You get an answer. “How many of this part do we have on hand?” “What’s the trend on this vendor’s lead times?” “Which jobs are running over?” The AI doesn’t just pull a number — it pulls context. It will often give you more than you asked for, in a useful way.
That’s not a gimmick. That’s a genuinely different relationship between a business owner and their own data. I’ve been describing it as “chatting with your data,” and I think that’s the right frame. The data was always yours. Now you can actually talk to it.
If you’re running Epicor Kinetic and you want to get Copilot configured against it, this is live now. Worth looking at.
Alright — that’s the week. I’m glad to be back, glad to be in a real office again, and very much looking forward to getting the team back around a table where we can actually think out loud together. More on the live stream schedule soon. Talk to you next week.
— Albert
Questions about anything we covered today? Schedule a free IT consultation and we’ll walk through it with your business in mind.