Every ERP vendor on the planet is talking about AI right now. “AI-powered insights.” “Intelligent automation.” “Copilot for everything.” If you’ve sat through vendor keynotes this year, you’ve heard the pitch. And if you’re like most manufacturers I work with, you’re still waiting for any of it to actually show up in your daily operations.
I get it. I’ve been implementing SAP Business One for manufacturers for years, and I’ve watched the gap between what AI promises and what it delivers inside an ERP stay stubbornly wide. Most of the AI features shipping today are surface-level — a chatbot here, a predictive widget there — nice demos, but nothing that fundamentally changes how your team works.
So we decided to stop waiting and start building.
At Third Wave, we’ve spent the last several months working with a startup partner to integrate a new AI-powered automation platform into SAP Business One — and what we’ve seen has changed how we think about AI in manufacturing operations entirely. This isn’t a product pitch. It’s an honest look at what’s actually working, what surprised us, and why we think the industry is approaching AI from the wrong direction.
The Problem with “AI Inside Your ERP”
Let’s start with the uncomfortable truth: most AI features embedded in ERP systems today are fundamentally limited by design.
Take SAP’s own AI Core, launching in the web client later this year. It’s a genuinely useful feature — an “Ask AI” button on every screen that lets you query data using natural language. But here’s the catch: due to data governance requirements (especially European data protection law), the AI can only see what’s currently displayed on your screen. It doesn’t have access to your full database. It can’t cross-reference your inventory against your purchase orders against your production schedule in a single query.
That’s not a knock on SAP — they’re navigating real regulatory constraints. But it means the AI can help you ask questions about what you’re already looking at. It can’t help you do work across your entire system.
For a manufacturer juggling thousands of SKUs, dozens of suppliers, and production schedules that change by the hour, that’s a meaningful limitation.
A Different Approach: AI That Builds Systems, Not Just Answers
Here’s where our thinking shifted. The conventional approach to AI in ERP is: take a large language model, point it at your data, and let users ask questions. It sounds great in a demo. In practice, it breaks down fast — especially with the messy, inconsistent data that lives in real manufacturing environments.
If you’ve ever copied a spreadsheet into ChatGPT and gotten back confidently wrong gibberish, you know exactly what I mean.
The approach we’ve been testing with our startup partner’s platform works differently. Instead of using AI to produce answers directly, it uses AI to write the code that finds, cleans, and interprets your data. It builds the system to answer the question — not the answer itself.
That distinction might sound academic, but in practice it’s the difference between an AI that hallucinates your inventory counts and one that programmatically pulls real numbers from your ERP, cross-references them, and gives you something you can actually act on.
What This Looks Like in Practice
Let me walk you through a few real scenarios we’ve tested. These aren’t hypotheticals — these are workflows we’ve built and demonstrated with actual customer data.
Turning a PDF Into a Quote in Minutes
One of our food manufacturing clients regularly receives ingredient lists from prospective customers as PDF documents. In the old workflow, someone on the sales team would manually look up each ingredient in SAP, check availability and pricing history, create a sales opportunity, and then build a quote. For a complex list, that’s easily an hour of work.
With the tool connected to SAP Business One, we fed it the PDF. It scanned the ingredient list, matched each item against SAP inventory, pulled historical pricing data, created the sales opportunity, and generated a linked quote with calculated pricing — all without anyone touching the keyboard. The whole process took minutes.
The sales rep’s job shifted from data entry to reviewing and approving the output. That’s a fundamentally different use of their time.
Processing Emails Into Transactions
We tested another workflow where the tool processes incoming emails containing purchase orders or invoices. It extracts the relevant data — line items, quantities, GL accounts — and creates the corresponding sales orders or AP invoices directly in SAP.
For manufacturers who process dozens of these documents daily, this isn’t a marginal time savings. It’s removing an entire category of manual work.
Bulk Operations That Used to Take Days
One of the more impressive demonstrations involved bulk data updates. We showed it processing a spreadsheet to create multiple new business partners in SAP, complete with payment terms and contact details. In another test, it updated bills of material across dozens of products — a task that would typically take someone days of careful manual work.
The key here is that it learns. When we told it to use warehouse-level costing instead of standard costs, it adapted its approach going forward. You train it once, and it remembers — much like onboarding a new team member who gets sharper with experience.
Why This Matters for Manufacturers Specifically
The data is messy. Years of manual entry, legacy migrations, inconsistent naming conventions, duplicate records — this is the norm, not the exception. Most AI tools choke on this. The platform’s code-generation approach means it can be trained to handle the mess rather than pretending it doesn’t exist.
The workflows are complex. Manufacturing isn’t a single-step process. A sales inquiry touches inventory, production planning, procurement, pricing, and shipping. AI that can only see one screen at a time can’t connect those dots. An AI that reads and writes across your entire system can.
The stakes are real. When you’re running a production line, a wrong inventory count or a missed purchase order isn’t an inconvenience — it’s a line-down event. Manufacturers need AI they can verify and trust, not AI that generates plausible-sounding answers they have to double-check anyway.
What We’ve Learned So Far
Start with repetitive, high-volume tasks. The biggest ROI comes from automating the work your team does over and over — document processing, data entry, bulk updates. Don’t start with “give me strategic insights.” Start with “stop making my team retype the same information into three different screens.”
The “exclusive channel” advantage is real. Third Wave is currently the exclusive SAP Business One channel partner for this platform. That matters because integration quality depends heavily on deep ERP knowledge. A generic AI tool that doesn’t understand SAP’s data model, business logic, and quirks will produce generic results. We’re building workflows that reflect how SAP B1 actually works in manufacturing environments.
AI as a “river guide” is the right mental model. One thing we’ve noticed working with customers is that the AI landscape is confusing. People hear “AI” and think ChatGPT. They see a copilot button in their software and assume it does everything. The reality is that different AI approaches are good at fundamentally different things. Our role isn’t just to implement tools — it’s to help manufacturers navigate what’s real, what’s hype, and what actually applies to their operation.
The integration story is just beginning. Right now we’re focused on SAP Business One connectivity, but the architecture supports connecting to anything with an API — SharePoint, CRM systems, email, you name it.
The Bigger Picture: What’s Changing in 2026
SAP’s web client is making real progress — version 2602 already covers 55–60% of desktop functionality, and they’re pushing toward 80% parity by year-end. The web client runs on any browser, any device, no remote desktop needed.
But the real story of 2026 isn’t the web client migration. It’s the emergence of AI tools that work alongside your ERP rather than being constrained by what’s built into it.
For manufacturers who’ve been waiting for “AI in ERP” to deliver on its promises, the better question might be: why wait for your ERP vendor to figure it out when the tools to build it yourself already exist?
What Comes Next
We’re currently spinning up trials with customers in our sandbox environment and building out common workflow templates — the kinds of automations that apply to almost every manufacturer on SAP Business One.
No slides. No hypotheticals. Just real data, real workflows, and an honest conversation about what AI can (and can’t) do for your business today.