AI has moved past the proof-of-concept stage for most manufacturers. The question is no longer whether it belongs in your ERP, but where it earns its keep and how to deploy it without exposing your data or chasing low-value use cases.
To dig into what this looks like inside Acumatica, we sat down with Doug Johnson, VP of Solution Architecture and Product Management at Acumatica. Doug leads product strategy for Acumatica’s cloud ERP platform and has been closely involved in the company’s AI roadmap since the start.
In this Q&A, Doug covers the two kinds of AI manufacturers should know about, how Acumatica protects your data when AI is in the mix, what anomaly detection looks like on the shop floor, what AI Studio unlocks, and where the roadmap heads next.
Doug: There are two kinds of AI worth knowing about.
The first is machine learning. A good example of this: someone emails you a bill as a PDF. The system reads it and pulls out the vendor’s name, the amount, and the date. If it gets the vendor wrong and you fix it, it learns — and next time, it gets it right.
The second kind is generative AI. That’s the more sophisticated. It can write, summarize, and create. Think of it as a new brain bolted onto your ERP. It works differently from a regular database, which is what lets it come up with answers in ways older software can’t.
Doug: This is a big one. If you’re not careful, AI can blow right past your security settings.
Someone could ask it, “Show me everyone’s salary,” and it would just hand that over. Acumatica has always controlled who can see what, and we’ve built those same rules into our AI. Every request — going in or coming out — gets checked against your permissions first.
If you’re not supposed to see something, AI won’t show it to you. If AI tries to write something weird into a product description, that gets blocked too.
Doug: ERP systems hold a ton of data, and the hard part is knowing where to look. Anomaly detection does that for you.
Every night, it scans your data and flags anything strange. So when your production manager logs in, they can immediately see where something is off. They can click into the issue, pinpoint the cause, and understand what happened. For example, an employee may have taken twice as long as usual at one station. Maybe it was their first day. You can leave a note, move on, and give them a chance to improve.
It works the same way for production orders. If actual cost runs far above planned cost, it flags the job so you can investigate right away instead of finding out weeks later. And it is not limited to production. It can spot issues across sales orders, purchase orders, and almost anything else in the system.
Doug: AI Studio lets you connect your own AI model to Acumatica. The reason that matters is that you stay in control. Your model lives inside your own walls, trained on your own data, and nothing leaks out. We support the popular ones — OpenAI (the company behind ChatGPT), Anthropic (the company behind Claude), and others. You plug it in, write a few instructions telling it what to do, and you’re off. It’s a bit more hands-on than the built-in stuff, but it gives you a lot more flexibility.
Doug: A few examples are built in. It can clean up an email — you type the rough idea, the AI makes it sound professional. It can read a support case and figure out how urgent it is. If a customer writes “I need this by tomorrow,” the AI bumps the priority automatically. It can write product descriptions.
Looking ahead, imagine job scheduling: feed the AI a pile of information about your shop and let it suggest the best order to run jobs in. That’s the kind of thing we think will be really powerful down the road.
Doug: Available today: reading PDF bills and receipts, anomaly detection, the interactive assistant, and AI Studio.
Coming soon: voice control — talking to Acumatica out loud and having it pull up data or build a chart for you.
Further out, we’re looking at things like predictive maintenance — using AI to predict when a machine is about to break down before it does — and smarter capacity planning. The goal is to let the machine handle the repetitive work so your people can stay focused on what matters most.
Three things guide how we build it:
→ Responsible — your data stays your data. We won’t ship anything that puts your security at risk.
→ Practical — there’s no point spending $40,000 on AI to catch a $1,000 mistake. The math has to work.
→ Easy to use — for the built-in stuff, you shouldn’t have to fiddle with anything. You flip it on and it works.
The bigger picture: whether it’s anomaly detection you can use today or voice control coming next year, the point is the same. Take the busywork off your team’s plate so they can focus on what actually moves the business forward.
This recap highlights the headlines, but the full webinar gets to the questions CEOs and CFOs are asking now: where AI can create measurable operational value, how to approach it without increasing risk, and what manufacturers should prioritize before making larger investments.
In the on-demand session, Doug goes deeper into anomaly detection using a live production order, walks through practical AI Studio prompts, and explains how Acumatica applies security controls across its AI capabilities. Joined by Tanner MacDonald from Aktion, the discussion also explores what manufacturers can do today to prepare their ERP environment for the next wave of AI adoption.
Doug Johnson has over 30 years of product management and marketing experience and is currently the Vice President of Solution Architecture and the Vice President of Product Management at Acumatica, where he defines business requirements for flexibly deployed ERP business management software using SaaS and Cloud technologies.
Christina Birmingham, Vice President of the Multi-Industry Division, leads the team responsible for delivering support, and services to companies in the Construction, Distribution, and Manufacturing Industries.
