Our team spent the SAP NOW AI Tour UKI in Birmingham talking to SAP users and business leaders on the stand, in sessions and over coffee. The demos were impressive. The conversations were more revealing, and one request came up again and again in different words: “Give me AI.”

The pressure usually starts at board level. Leadership teams have seen what AI can do and want to know what their business is doing about it. They also want to know what return it will bring. Many of the people we met are stuck between those two demands, convinced AI matters but unsure whether they are ready or where the first step should be.

As our Commercial Director Richard put it after the event, “Give me AI” is a real strategy problem. Here is what we heard, and what we think organisations running SAP should do about it.

The Question Has Changed

A year ago, most AI conversations started with “Can we build it?” At the NOW AI Tour, that was rarely the first question. The people we spoke to wanted to know what AI will cost to run, how it will be secured, who will govern it and how it will keep up as the technology changes.

Jakob, from our commercial team, has seen the same shift in almost every conversation. “One hundred per cent. The focus has moved to cost, security and governance. The biggest driver is recent concern about confidential business information and AI. People want to know exactly where their data goes and who can see it before they commit to anything.”

That is a healthy change. Asking about running costs and governance early means AI gets treated as a long-term capability with an owner, a budget and rules, which is how it delivers value over time.

The Three Things Holding Organisations Back

When we asked Jakob which worries he heard most, the same three came up again and again. Each one is real. None of them is a reason to wait.

“We don’t have use cases”

Often this is less about a lack of ideas and more about where people are looking. The first list most organisations write is full of efficiency tasks: speeding up a manual process, clearing a backlog, cutting admin. Those are worth fixing, but they rarely need AI. Better processes or standard automation usually solve them faster and for less.

The use cases worth chasing are the ones where AI changes the outcome itself: spotting a risk before it lands, predicting demand early enough to act on it, or making a decision with more information than any person could weigh in the time available. Start by asking what you could achieve that simply isn’t possible today, then work back to the data and processes that would make it happen. A structured AI use-case workshop is a good way to find those opportunities and turn them into a shortlist.

“We’re not on S/4HANA yet”

Many of the AI features embedded in SAP’s applications are built for S/4HANA, so this concern makes sense. It doesn’t have to stop you. SAP’s newest AI capabilities, from building your own agents to working with large language models, are delivered through SAP BTP, and they’re available whichever version of SAP you run. That means you can start building and proving value with the latest technology today, while the move to S/4HANA is still being planned.

“Our data isn’t ready”

This is the most self-aware of the three, and the most important. Gartner predicts that through 2026, 60% of AI projects will be abandoned because of insufficient data quality. AI acts on whatever data it is given, so if customer, supplier or product records are wrong, the results will be too.

It is also where most conversations at the NOW AI Tour began. “Most people come to us with a use case, but the conversation always starts with the foundations,” Jakob told us. “Is your data in a healthy enough state for AI to work?” Our data quality guide explains how to find out.

The AI Business Case Needs to Change

Most technology business cases follow a familiar formula: cut this cost, speed up that process, save a set amount each year. That works well for many AI projects. Automating invoice matching or ticket triage has a clear, measurable return.

Richard’s point is that some of AI’s biggest value is much harder to put a number on. What is it worth to spot a supply problem a week before it hits? To avoid a compliance issue that never happened? To make a pricing decision days before a competitor does? These outcomes are real, but they rarely fit neatly into a spreadsheet. If investment decisions only approve what can be measured precisely today, organisations risk passing on the capabilities that could set them apart tomorrow.

That doesn’t mean abandoning financial discipline. In practice, we see three approaches work well:

• Fund in stages. Start with a small prototype that proves the idea, then commit more once you have evidence.
• Track leading indicators. Measure the things that come before the financial result, such as how much earlier a decision is made, how many exceptions are caught automatically or how much time teams get back.
• Cost the full lifecycle. Build the running, security, governance and maintenance costs into the case from day one, so the numbers still hold a year after go-live.

Where to Start: Foundations, Then a Prototype

The most encouraging moments at the NOW AI Tour came when someone’s thinking shifted from “we’re not ready” to “we could actually do that.”

“The moment that changes things is when someone realises their use case is far more realistic than they thought,” Jakob said, “and that we can prove it with a rapid prototype.”

These are the steps we recommend:

1. Check the foundations. Look at the data the use case depends on, and fix the gaps that would undermine the results.
2. Pick one problem. Choose a use case with a clear owner, a real business outcome and data you can trust.
3. Prototype quickly. Build a working version in days, test it with the people who will use it, and learn what it would take to run it for real.
4. Decide with evidence. Use what the prototype shows, including running and governance costs, to decide whether to scale.

A prototype costs a fraction of a full project, and it turns a debate about AI into a decision based on something people have actually used.

From “Give Me AI” to Real Results

The organisations that will get the most from AI are those that know where it can change an outcome, understand what it will cost to run, and are ready to take the first step. Being first to the technology matters far less.

That changes the question boards should be asking. “What can we do with AI?” invites a long list of possibilities. The more useful question, and the one Richard left the event with, is which outcomes become possible with AI that simply weren’t possible before.

If your organisation is facing its own “give me AI” moment, we’d be glad to help you find the answer. Our AI use-case workshop helps you check your foundations, shortlist the right use cases and plan a first prototype, or you can talk to our team directly.