Article | October 28, 2026

Stop asking whether you need an AI agent

A practical approach to helping SAP customers determine when AI agents are the right solution and when a simpler approach may deliver greater value.

By Roberto Simi, Global SAP Business AI Offering Leader, DXC Technology

Over the last year, DXC has had countless conversations with SAP customers about AI.

The discussion usually starts in the same place: agents. Which agents are available? Which agents should we build? How quickly can we deploy them?

What we’ve seen is that many organizations are becoming increasingly confused, not because there isn't enough innovation happening in AI, but because there's almost too much of it. Every week brings another announcement, another capability, another roadmap. We've moved from AI features to copilots, from copilots to assistants, and now to agents. Somewhere along the way, many businesses have lost sight of a more fundamental question: what problem are we actually trying to solve?  

The challenge facing most SAP customers today isn't building an agent. 

The challenge is understanding whether an agent is the right answer in the first place. The technology is advancing at an incredible pace, but that doesn't mean every business problem requires an AI-driven solution. In many cases, organizations are focusing on the latest capability before they've properly assessed the underlying business need.

One of the reasons this may be happening is that we've seen a similar story before. Many organizations invested heavily in robotic process automation with the expectation that automation would transform the enterprise. The technology itself worked, but the long-term reality was often more complicated. Bots had to be maintained, systems changed, business processes evolved, and over time the cost of keeping automations running became a challenge in its own right. The lesson from RPA is that the technology itself isn’t necessarily the issue. Maintenance became the hidden cost nobody fully anticipated.  

Agentic AI introduces many of the same considerations, but with an important difference. When an RPA bot failed, everyone knew about it. It stopped working. An AI agent can continue operating while producing outputs that appear entirely credible but are fundamentally wrong. That's why governance, monitoring and measurement are so important. The conversation can no longer be limited to what an agent is capable of doing. It must also include how that agent will be managed and evaluated over time. RPA failed loudly. Agents fail quietly. 

This is also why the industry's focus on "build versus buy" can miss the point. The more important question is ownership. Every AI capability sits somewhere along a spectrum. At one end are capabilities already delivered and maintained by SAP. At the other are fully bespoke solutions that an organization is responsible for governing, maintaining and evolving. Every step along that journey introduces additional complexity, additional responsibility and, ultimately, additional technical debt. The goal shouldn't be to build the most sophisticated AI solution possible. It should be to identify the lowest ownership option that can reliably solve the business problem.

When customers ask where they should begin, we come back to three simple questions. First, can you name the pain with evidence rather than opinion? Second, is the problem genuinely agentic in nature, or is there a simpler form of automation that could achieve the same outcome? Third, can SAP already provide the capability you're looking for? These questions sound straightforward, but they often lead organisations to very different conclusions than they expected when they started the conversation.  

The second and third questions are where the discussion usually becomes interesting. Most organizations arrive expecting to talk about custom AI agents, but many business processes aren't truly agentic. They are structured, repeatable and governed by known business rules. In those situations, a custom-built agent may introduce complexity without creating additional value. In fact, some of the best outcomes come from discovering that the problem can be solved through a capability that already exists within SAP, or through configuration rather than custom development. As SAP continues to expand its portfolio of AI features, assistants and agents, the challenge is often less about what can be built and more about understanding what has already been built for you.  

The answers to these questions also help determine how much ownership an organization should take on. Every time you move from consuming a standard SAP capability to building a bespoke agent, you take on more responsibility for governance, maintenance, monitoring and long-term evolution. Sometimes that investment is entirely justified because the use case is unique and provides meaningful competitive advantage. But many organizations are rushing towards custom AI development before they've exhausted the capabilities already available to them. In a world where the underlying technology is changing at an extraordinary pace, that can be an expensive mistake.  

Ultimately, the organizations that will realise the greatest value from AI won't be the ones that build the most agents. They'll be the ones that consistently route each problem to the lowest-ownership solution that reliably solves it. Sometimes that will be a custom agent. Sometimes it will be a configured solution running on SAP's managed platform. And sometimes it will simply be an existing SAP capability that was already sitting there waiting to be activated. The organizations that succeed won't be driven by the excitement of building AI. They'll be driven by the discipline of choosing the right AI.

The key message for SAP leaders is this: stop starting with the technology.

Instead, bring one real business challenge. Ideally, make it one that's measurable and one that's attracting attention from your CFO, your operations team or your business stakeholders.

Then ask three questions:

• Can we prove the problem exists?
• Is this genuinely an agent problem?
• Does SAP already provide a capability that solves it?

The answers might lead you towards a custom AI agent. They might point to a configured solution on SAP's managed platform. Or they might reveal that the capability already exists within your SAP landscape.

Either way, you'll have something far more valuable than an AI strategy.

You'll have clarity.