ERP vendors have long sold the idea that their prepackaged processes represent the pinnacle of efficiency. Orders-to-cash, procure-to-pay, plan-to-produce—they’ve all been codified, standardized, and stamped with the industry best practice label. But there’s a problem: these classic process models assume that people live inside the ERP system, dutifully clicking through transactional workflows and running list reports to see what to do next.
However, if you walk the halls of any office or shop floor, you’ll hear a different story. Deals are negotiated over email. Exceptions get handled through chat messages. Forecasts and approvals live in spreadsheets. And in many industries, the final “processing steps” are still scribbled down onto paper forms.
As the official system of record, ERP systems still play a vital role within the business. However, if we're being honest with ourselves, it’s oftentimes not where the day-to-day work actually happens. The real work takes place in the handoffs, the conversations, and the improvisation that fills the gaps between those neatly defined ERP transactions. That’s where inefficiencies creep in, where “shadow IT” emerges, and where business users start inventing their own ways to get things done.
If we’re serious about process optimization, we can’t just tinker around the edges of ERP workflows. Instead, we need to zoom out and reimagine how processes flow across the entire digital workplace. With that in mind, this article explores what it looks like to confront these gaps head-on and how rethinking the model can lead to processes that are not only “best run” on paper, but truly best run in practice.
Shadow IT as a Symptom, Not a Sideshow
Enterprise IT departments often dismiss shadow IT as little more than a handful of desktop apps or rogue Excel spreadsheets flying under the radar. If these makeshift solutions can plug some nuanced gaps here and there and make department heads happy, then IT is content to look the other way on these small-time skunkworks operations.
IT benevolence aside, this dismissive attitude is rooted in a broader overconfidence: the belief that ERP and other business systems provide comprehensive coverage for core business processes.
Ask an enterprise architect, and they’ll often speak confidently about their ERP system’s ability to fully support processes such as orders-to-cash, procure-to-pay, or plan-to-produce. But when we sit down with end users, we usually hear a very different story.
Sure, these users spend a lot of time working in ERP systems recording transactions. But the real work—the conversations, the heavy thinking, the collaboration, the back-and-forth—that's happening elsewhere. Frequently, this work takes place over email, chat, phone calls, or even paper forms. And increasingly, users are leaning on agents like Microsoft Copilot or ChatGPT to shoulder some of the cognitive load, helping them analyze information, draft responses, or make informed decisions.
Figure 1: Working with AI-Powered Research Agents
No matter the channel, the point is that this external activity is a clear signal that the system isn’t giving users the level of support they need to actually complete their work. Sometimes it’s due to missing functionality. For example, certain industries simply don’t align very well with “standard” ERP processes. Other times, it’s because the system-defined process flow runs contrary to how the work naturally unfolds. This is particularly true for all-or-nothing transactions which assume that all the details are known at once, when in reality those details often take time to come together through collaboration across teams.
The bottom line? Show us a monster-sized Excel spreadsheet or an Access database packed with VBA code, and we’ll show you an area where your ERP system doesn’t cover nearly as much ground as you think.
Why AI Sprinkles Won't Cut It
As the AI arms race continues to heat up, ERP vendors are racing to deliver some tangible AI wins. In many cases, these innovations are being delivered in the form of sidecar-style enhancements that improve user experiences within the ERP application itself. Think about features such as automated field suggestions, smarter search, or AI-assisted document handling. These new capabilities are designed to make life a little easier for users, reducing friction and creating a sense of momentum.
While these new features certainly add value, they don’t get at the heart of the issue. The fundamental problem is that ERP systems still expect you to come to them. The system remains the destination, and users have to bend their workflow to fit its structure.
Take procure-to-pay as an example. With today’s “AI sprinkles” approach, your ERP might offer a feature like automatically extracting line items from an invoice or flagging unusual payment terms. Helpful? Sure. But the broader process still assumes that approvers drop what they're doing, go log onto the ERP system, navigate to the right screen, and perform an approval step to push the workflow forward.
Now, let's imagine a different approach that better aligns with the actual flow of work. Instead of waiting for users to pull transactions into ERP, the system proactively brings the work to the appropriate user(s) at the appropriate time. For example, a procurement manager could receive a Teams message summarizing an outstanding purchase order, complete with supplier history and AI-generated risk insights right there at their fingertips. Then, right within Teams, they can approve the request, request clarification, or even trigger follow-up actions without ever leaving the conversation.
Figure 2: Processing Invoice Approvals within the Flow of Work
Orders-to-cash presents a similar story. An ERP might surface AI-powered credit checks or auto-suggest payment terms, but the process still leans heavily on emails, phone calls, and spreadsheets to chase down late payments or confirm order details. A reimagined model would orchestrate those steps across channels: an AI agent could automatically draft a collections email, notify sales when a customer’s order is at risk, or deliver a summarized account status directly in Outlook or Teams. Again, ERP serves as the system of record, but the real work flows through the tools people already use every day.
That’s the difference between sprinkling AI on top of legacy processes and rethinking how work should actually get done. To unlock true decision support, intelligence must be embedded into the flow of work so that we can deliver the right information to the right people at the right time.
Better, Stronger, Smarter: Moving Towards Systems That Think
We can rebuild it; we have the technology.
Up to this point, we’ve talked about some of the gaps in ERP functionality and where real work tends to spill outside the system. But let’s be clear, this isn’t one of those “ERP is dead” posts. Far from it. ERP systems still deliver tremendous value as the backbone of business operations and as the system of record for critical transactions.
The point is that ERP on its own isn’t enough. The gaps we’ve highlighted don’t signal that ERP has failed. Instead, they highlight the need for an additional layer of intelligence that guides users through process flows as they actually unfold. By weaving data and AI into the flow of work, we can move beyond simple record-keeping and create systems that actively support decision-making and collaboration. In other words, ERP continues to play its central role, but we need to complement it with a intelligent nervous system that makes the whole operation run more effectively.
Building a Digital Nervous System
As we've seen, ERP systems operate on an assumption that users will come to them. As a result, they tend to take on more of a passive role in the process, waiting for someone to log in, run a report, and act on the information.
To put this concept into perspective, consider the list report shown in Figure 3 below. Here, a production controller might use this report in SAP to keep tabs on in-flight production orders, investigate delays, and respond to issues. The system faithfully presents the data, offering a few actions/links to respond to issues, etc. However, the functional support basically stops there. The rest of the burden falls on the user to sift through rows of information, spot patterns, and decide what actions to take.
Figure 3: Monitoring Workflow Using List Reports
This is the limitation we need to address. Instead of expecting people to constantly draw insights out of static reports, we need to introduce a digital nervous system over the top of the ERP system to keep track of what’s happening and, more importantly, proactively respond based on available information.
Figure 4 illustrates what a digital nervous system like this could look like building on Real-Time Intelligence in Microsoft Fabric. Starting from the top of the figure and working our way down, you can see that the ERP system is one of many data sources being fed into an event stream that drives this intelligence engine. This unified stream of data makes it possible to keep a finger on the pulse of the goings-on across the entire enterprise (and beyond).
Figure 4: Moving from Reactive to Proactive with Fabric Real-Time Intelligence
Once the event stream is funneled into a highly-performant event house in Fabric, we can then put that data to work in two compelling ways:
Real-Time Dashboards: We can feed the data into real-time dashboards to provide a more comprehensive view of what's going on in a particular business area. From a data perspective, this can be a huge win as similar reports in ERP systems usually don't have visibility to what's going on outside of the ERP system (e.g., what's happening on the shop floor, where maintenance technicians are currently located geographically, and so forth).
Workflows & Automation: If we assume that the event house has access to up-to-date information from across the enterprise, it follows that we can apply custom business rules to respond to certain events by triggering workflows. Within these workflow processes, it's a bit of a choose your own adventure scenario in terms of how the system responds.
While real-time dashboards are compelling, our focus for this article is more around the workflow and automation capabilities. Looking back at Figure 4, you can see how the business rules we define control something called an activator which acts as a kind of a neurotransmitter for routing events into workflow agents that can take proactive action. In some cases, that action might be as simple as sending an alert email to relevant process stakeholders. However, in many cases, the system has all the information it needs to resolve the issue on its own without the need to loop in a human being.
Rethinking Process Design
From a flow perspective, the real-time workflow capabilities outlined in the previous section are huge because they allow us to completely re-think the way that business processes are designed.
Normally, whenever we sit down with teams to explore what “better” might look like, they struggle to imagine anything different. Years of working around technical constraints have shaped their expectations, and there's a real struggle to envision process flows that go beyond the limitations they have become accustomed to.
However, this mindset changes dramatically when we collectively shift from thinking about the system as a passive tool that simply records transactions to an active participant in the process, one that works alongside humans rather than waiting for them to call it to action.
This shift unlocks many new possibilities. For example, we can extend the reach of the system by hosting AI-powered agents that operate like collaborative team members. These agents can be tasked with handling routine steps, surfacing insights, or even suggesting next actions in real time. Here, it's not just about automating tasks, it is about embedding intelligence into the process itself so the system learns and adapts as the business evolves.
With these capabilities, organizations can:
Train the system to follow business processes by encoding workflows and policies that reflect how work really happens.
Build AI models for advanced decision support, helping users evaluate tradeoffs, identify risks, and make smarter choices.
Deploy AI-powered agents that engage directly with users in the flow of work, whether in Teams, Outlook, or mobile apps, so guidance comes to them rather than the other way around.
To put these concepts into perspective, consider the process of onboarding a contingent worker. In a traditional model, HR might enter basic details into the ERP system, then manually chase down IT for system access, facilities for a badge, and a manager for equipment approvals. In a reimagined model, AI agents coordinate these steps automatically. As soon as HR initiates the process, the system can notify IT of access requirements, generate tasks for facilities, and even pre-populate an equipment order for the manager to review. The ERP remains the system of record, but the flow of work is orchestrated intelligently across departments.
The beauty of this approach is its flexibility. You do not have to transform everything at once. You can innovate at your own pace, tackling broken processes one at a time. And because this innovation happens around the edges of the ERP system, there is no need to wait for a costly upgrade cycle.
By introducing an intelligence layer, we allow ERP to focus on its strengths as the system of record while we innovate around the edges in new and exciting ways. Reimagined user experiences, powered by AI agents and smarter workflows, can deliver the kinds of enhancements people often associate with major ERP upgrades, without the disruption or expense.
Closing Thoughts
ERP systems remain a critical foundation for running the business, but as we have seen, much of the real work happens outside of those systems. Shadow IT, workarounds, and manual coordination are not signs of resistance, but signals that users need better support in the flow of work. Addressing these gaps does not mean discarding ERP. It means complementing it with a layer of intelligence that helps people collaborate more effectively and make better decisions.
By rethinking process design with AI-powered agents, data-driven insights, and more adaptive workflows, you can evolve your systems from passive record-keepers into active participants in daily operations. And as long as everyone's on the same page architecturally, you can really innovate at your own pace. Small, targeted improvements can add up to meaningful progress, allowing ERP to play to its strengths while giving users the flexibility and guidance they need to optimize process flows end-to-end.
