ERP in 2024: What the Next Wave Looks Like

When I started Arvension three years ago, ERP modernization meant APIs and cloud migration. Then last year, it shifted. Now every major player—SAP, Oracle, Workday—is announcing AI capabilities. But here's what nobody's saying out loud: most of these are bolted on. They're add-ons to systems designed in the 1990s.

2024 is different. This is the year enterprises stop asking "how do we add AI to our ERP?" and start asking "what if we rebuilt ERP with AI at the core?"

The Infrastructure Problem That AI Solves

Most enterprises are running ERP systems that were designed for a completely different world. When SAP R/3 was built, the problem was data capture and workflow automation. Process Manufacturing happened here. Requisitions got approved there. The system's job was to move data through predefined paths.

But modern business doesn't work that way anymore. You have real-time supply chain data, customer signals coming from six different systems, market conditions changing hourly. Your ERP is still moving data through the same predefined paths. That's not a feature anymore—that's a liability.

AI changes this fundamentally. An AI-native ERP doesn't just execute workflows; it understands context. It can ingest unstructured data from suppliers, market feeds, customer interactions. It can reason about what decisions need to be made and surface them to humans, not buried in a dashboard nobody opens.

Last month, we worked with a mid-sized manufacturer who was drowning in their ERP. They had 40 different approval workflows. Most requisitions were getting stuck waiting for approvals for reasons that weren't even in the system (a buyer knew something wasn't needed, but couldn't express it in the form fields).

When we injected an AI understanding layer into their approval process, overnight that number went from 40 stuck approvals a day to maybe 3. The system understood exceptions. It understood context. It knew when to escalate and when to just route automatically.

What AI-First ERP Actually Looks Like

The next generation of ERP won't look like Salesforce with an AI chatbot. It will look completely different.

First, the data model becomes simpler. Right now, ERP systems have 15,000 tables and you spend months mapping your business into them. AI-first systems can handle messiness. You feed the system your raw data—from your supplier portal, your warehouse management system, your customer communications—and the AI creates the internal model. You stop living in the ERP's rigid schema and let the ERP adapt to you.

Second, decision-making becomes intelligent. Today, your ERP system is a database that follows rules. Someone set the reorder point at 100 units. It stays 100 units. An AI system doesn't work that way. It learns patterns. It sees that you actually need 150 units in February because of seasonal demand, but only 50 in August. It adapts.

Third, interfaces become conversational. Nobody wants to fill out seven form fields to process an expense. They want to say "I bought supplies for the warehouse, $340" and have the system figure out the rest. Most of that figuring is just AI-powered classification and routing.

The Integration Headache You Don't Have Yet

Here's what I'm not seeing enough people talk about: if you rebuild ERP around AI, your integration nightmare gets worse before it gets better.

You still have legacy systems. You still have a supply chain you're not replacing wholesale. An AI-first ERP has to sit in the middle and make sense of all that chaos. That's a harder problem than building the ERP itself.

But it's solvable. The enterprises that figure out the integration pattern here—not the technology, but the actual pattern for safe, iterative integration—those are the companies that will run circles around everyone else in 2024.

What You Should Do Right Now

If you're running SAP, Oracle, or Workday: don't wait for the vendor to give you AI capabilities. You already have the data. You already have the processes. Start experimenting with an AI understanding layer on top. See what becomes possible when you can ask your ERP questions, not just run reports.

If you're building new systems, you know what to do: put the AI model before the database schema. Let the AI define the structure, then make the structure flexible enough to adapt.

The next wave of ERP isn't about better dashboards or faster deployment. It's about systems that actually understand your business. The infrastructure to make that happen is finally here. What you do with it in 2024 is going to separate the companies that move and the companies that stall.