I’ve been thinking and talking recently about the risk of approaching SAP transformation with a Maslow’s Hammer mindset. That is, treating every business challenge as if it demands a massive, all‑encompassing transformation program. Some feedback I’ve gotten included a practical question: “What are the real AI business use cases companies should be pursuing today?”

The short answer is this: Many of the most valuable AI use cases can and should be deployed before large‑scale ERP transformation, not after. This article is my attempt to explain why.

The directive to “use AI” has become nearly universal. Most of us already rely on it for day‑to‑day administrative tasks, and at Lemongrass, we’re leveraging AI across operations, support and even project execution. But we’re a technology‑centric company. Many large enterprises operate with very different levels of complexity, legacy and organizational inertia.

Given our roots, we naturally focus on SAP‑related AI use cases. The long‑term vision in which AI is embedded ubiquitously across every layer of a business process, quietly optimizing, predicting, and automating, is compelling. Hardening supply chains, improving demand forecasting, maximizing manufacturing output and increasing revenue with minimal incremental cost of sale represent enormous opportunities. It’s no surprise that every boardroom is talking about it.

Where many organizations struggle, however, is the assumption that achieving this future state requires a sweeping ERP transformation first.

Yes, clean, modern, well‑integrated ERP landscapes make AI easier. No one disputes that. But meaningful, high‑value AI opportunities exist today, without a multi‑year S/4HANA migration, a greenfield reset or a full‑scale master data overhaul.

In reality, many organizations already have enough foundational “plumbing” to unlock tangible AI‑driven value. These initiatives can deliver benefits quickly and, importantly, help fund and inform the more complex transformations that follow.

Take exception management as a simple example. Many enterprises employ large teams to handle invoice errors, purchase order mismatches, policy deviations, missing approvals and incorrect subtotals. This is an ideal GenAI use case, and one that can be implemented far faster than most expect.

It does not require a full‑blown transformation or a “let’s start again” approach. It does not demand perfect data or a multi‑year MDM program. With the right architecture, the majority of these exceptions can be resolved in seconds, without human intervention, fully integrated with SAP and updating transactional data in real time.

Some exceptions exist because of flawed processes, of course. But if technology can resolve them instantly, does the process truly need to be redesigned from scratch?

One of our clients applied a similar philosophy by combining multiple sources of customer‑intent data with real‑time inventory visibility. By deploying a GenAI‑powered sales agent to identify missing products and services on invoices, they unlocked hundreds of millions in incremental top‑line revenue. Again, fully integrated and automatically updating orders directly within their cloud ERP systems.

These are pragmatic, high‑impact use cases that can be adopted quickly. Yet many organizations are still convinced they must first complete large, capital‑intensive ERP programs, migrate off legacy systems, cleanse all master data, or replace entire application landscapes before they can begin capturing AI‑driven value.

This is where the methodology itself deserves scrutiny.

A transformation‑first mindset treats AI as a downstream reward, something earned only after years of disruption, cost, and risk. An AI‑first mindset, by contrast, treats AI as a diagnostic and value‑creation tool: one that exposes where data truly breaks down, which processes genuinely cannot be automated and where transformation effort is actually justified.

While an AI‑first approach is not without risk, the pace of innovation and AI’s rapidly expanding ability to build, retain and reason over contextual information should give pause to organizations about to embark on multi‑year transformation programs. When will AI be able to consume and traverse interconnected business processes end‑to‑end? Probably sooner than many expect. Not because enterprises are perfectly aligned but because AI increasingly infers those connections from observed behavior rather than prescribed design. Of course, this won’t solve the meaty problems of governance and getting the business aligned, but if the size of the big ‘T’ transformation is reduced, even this gets more manageable.

In a perfect world, perhaps transformation would always come first. But we don’t live in a perfect world. We live in one that is changing and accelerating, faster than ever.

The real question is this: If AI can be deployed rapidly across ERP and the broader ecosystem to solve material business problems, generate measurable value and reveal where transformation is genuinely required, shouldn’t our methodologies start there?

Today’s dominant thinking tends to run counter to this. One could argue that this conveniently aligns with the interests of large integrators and technology vendors, bringing us full circle to Maslow’s Hammer problem.

There is enormous value available right now that can be captured in manageable, low‑risk slices. And if AI can fund your transformation, sharpen its scope and reduce its risk, the question isn’t whether you can start with AI… It’s why you wouldn’t.