AI Bolted On or AI Built In? What Every Business Should Ask Its ERP Vendor

For many small and midsized businesses (SMBs), using an ERP solution with a general-purpose artificial intelligence (AI) tool seems like a reasonable starting point. Many of the leading LLMs are capable, well-supported, and already familiar to a lot of teams and integrating the AI is not hard – particularly if you have open application programming interfaces (APIs). What is hard is relying on it for the kinds of questions your business actually needs answered.

In this, the third blog in our AI thought leadership series, (AI on Your Terms: Managed and Configurable AI and Why the Distinction Between AI and Automation Matters for ERP), we continue to explain some core concepts every business should know about AI and provide some helpful definitions of terms and concepts.  Today we focus on the reasons why SMBs relying on an ERP solution with an external AI will quickly discover that this “bolted on” approach is not the same thing as having AI features and capabilities “built in.”

How a General-Purpose AI Really Works

Let’s jump right in.

An external AI connected to an ERP solution receives data on request. It gets a snapshot of information: the records available at the moment the query runs. But you and your team need the full picture.

ERP data carries meaning through its relationships, its workflow state, and the logic the system applies to it. An invoice belongs to a purchase order, sits in an approval chain, and has a status that reflects where it is in your process. As it’s designed to do, an external AI tool sees the fields but understanding what those fields mean requires knowing the system they live in—and the bolted-on AI doesn’t have that understanding, creating a gap.

That gap shows up in questions that seem straightforward from the outside. For example, “Which invoices are pending approval?” sounds like a simple query. Answering it correctly requires knowing where each invoice sits in the approval workflow, which approvers have acted and which have not, and what conditions need to be met before the next step can proceed.

External AI can retrieve data. Connecting that data to the workflow logic it lives inside is a different problem.

Bridging the Gap: Configuring Your AI Tool or Choosing an AI-Powered ERP

The answer to the external-AI problem is not to abandon these integrations but to be clear about what they are from the start.

A general-purpose AI tool connected to an ERP solution handles certain tasks well: summarizing content, answering questions that do not require deep system context, or pulling surface-level data on request. But for operational questions that require workflow awareness, approval chain logic, or an understanding of how your ERP entities relate to each other, the gap grows quickly.

You can configure an LLM to understand your approval chains, entity relationships, and workflow states, but that configuration work is substantial. At a certain point, you’re not extending your ERP system with AI; you’re rebuilding ERP logic inside an AI tool, at real cost in time and ongoing maintenance.

Building AI for ERP software from the ground up means designing it around the data model, the workflow logic, and the relationships that give ERP data its meaning, rather than reconstructing all of that from the outside. The result? Practical AI with powerful results.

Acumatica, with our AI-first product strategy, is embedding practical, secure, and scalable AI into our ERP technology. Our approach is not to bolt on AI tools and capabilities to outdated ERP systems, but to build these capabilities into the workflows of business management technology so it works seamlessly with your business processes. Our AI Assistant operates directly within the ERP, with access to live workflow state, entity relationships, and the system context that makes data meaningful. That context is what allows it to answer operational questions reliably, not just surface the data behind them.

As you consider your options, a useful benchmark for any ERP AI capability is this – does the AI know your ERP, or does it know about it? The answer to this question will shape what you can reasonably ask it to do.  We will continue to answer this important question in the next blog in our ‘AI That Works for You’ series, Purpose-Built AI Beats Horizontal AI.

Until then, you can learn more about Acumatica’s approach to AI by watching our insightful, on-demand webinar, Acumatica AI: Transforming ERP for the Intelligent Business Era, and you can also let us know if you have any other AI in ERP questions.