AI, SAP and the future of Quality Engineering

AI is beginning to reshape both software development and Quality Engineering. But for SAP organisations, where complex end-to-end processes underpin critical business operations, speed alone is not enough. We spoke with Roman Zednik, Field CTO EMEA at Tricentis, about how AI is entering the SAP landscape, why business context is becoming essential to testing, and why Quality Engineering could ultimately become a C-level concern.

How is AI changing SAP Quality Engineering?

How do you expect AI to change the SAP landscape over the next three to five years?

Roman Zednik: SAP customers are in a somewhat different position from organisations working mainly with custom applications. Many are still heavily focused on major transformations, moving from the classic SAP world to new environments, whether public or private cloud. These are business-critical processes, so it is understandable that SAP organisations tend to be a little more conservative in adopting new technologies.

What I see now is AI being embedded step by step into end-to-end business processes. A process may still consist largely of traditional IT components, but at a certain point an agent takes over one part of it. I don't expect complete SAP processes to suddenly become 100% agent-driven, certainly not in the next one, two or three years. But agents will increasingly become part of those processes. 

If AI increasingly creates software, should AI also be trusted to validate it?

Roman Zednik: AI supports both development and testing, but there needs to be independence. The same agent should not write the code, write the test case, execute it and then analyse the result.

It is really the same principle we have always applied with people. The developer shouldn't simply validate everything they have created themselves. AI doesn't change that basic principle. 

Why AI changes the challenge of SAP Quality Engineering

Does faster AI-driven development create a new bottleneck for Quality Engineering?

Roman Zednik: We are already seeing this. Development teams generate much more code than before, including AI-generated code. But somebody still has to quality-assure it. So suddenly the QA teams have much more to test.

Customers are asking us how they can adapt Quality Engineering to the increased speed of development. Technology is part of the answer, but methodology becomes even more important than ever before. 

There is another dimension that is becoming increasingly important: cost. Initially, using AI was relatively inexpensive and in some cases practically free. Now organisations are beginning to see how quickly token consumption and other AI costs increase. So AI will make you faster, but that does not automatically mean it makes you cheaper. 

Why is business context so important when applying AI to SAP Quality Engineering?

Roman Zednik: Generating a test case with AI is relatively easy. The much more important question is whether it is a useful test case. Does it understand the enterprise and the business? Does it fit into the complete end-to-end process? Is it actually testing something business-critical?

That is particularly important in SAP because a lot of knowledge still sits with subject matter experts, business users and business owners. You may have somebody with 20 or 25 years of SAP experience who understands how processes connect and what test data really matters. An AI agent doesn't automatically have that knowledge. 

The syntax of AI-generated code will continue to improve, and security is increasingly incorporated as well. But the semantic question, the business context, is the difficult part. Systems often fail at the points where processes and systems integrate, not because an individual component doesn't work. 

What is agentic Quality Engineering, and how does it work?

Is that where Tricentis' acquisition of Tabnine comes in?

Roman Zednik: Yes. One of the main reasons for acquiring Tabnine was its technology around context. The idea is to build a context model from the knowledge available within an enterprise, from sources like service tickets and knowledge bases. That gives the AI a much better understanding of the company and its business processes.

We then combine that context with agents that generate tests in Tosca or qTest. The goal is that the agent doesn't behave like someone who has just arrived in the company. With the right context, it should behave much more like an experienced person who understands the environment and knows what is important. 

It should also help with AI consumption. If the relevant enterprise context is available proactively rather than having to recreate all of that context through prompting every time, you can reduce the amount of AI consumption required. 

Where does SAP Signavio fit into this picture?

Roman Zednik: Signavio is a very valuable source of business-process context. You can model the business processes in Signavio, bring that through the SAP toolchain and connect it with Quality Engineering and test automation. That means you relate the most important business processes to test coverage and ultimately bring the test results back into that process landscape. 

SAP has also changed the terminology here. What was previously referred to as the SAP Integrated Toolchain is now called the Agent-led Toolchain. That reflects the increasing role agents will play across these connected tools. 

But there is an important condition. If you use Signavio to model your processes, you have to maintain those models. We've seen organisations create very detailed process models and then stop updating them while the software and processes continue to change. At that point the model becomes outdated and loses much of its value for Quality Engineering.

What does Agentic Quality Engineering look like in practice today?

Roman Zednik: One foundation is MCP. We have added MCP server interfaces to core Tricentis products including Tosca, qTest, NeoLoad and SeaLights. I sometimes compare MCP to USB-C. It provides a standardised way for agents to communicate with tools and with other agents. 

On top of that, we have agents for specific Quality Engineering activities. In qTest, for example, an agent takes a requirement or specification and generates test cases. On the performance side, an agent analyses load-test results and infrastructure information to help identify the cause of a performance problem. 

We also have Tricentis AI Workspace, where organisations create workflows involving different agents and systems. You might identify a high-priority ticket, analyse the business context, check whether test cases already exist and generate them if necessary. 

SAP has been an important focus from the beginning. When we developed our agentic test automation capabilities, SAP was the first technology area we focused on, including SAP Fiori and SAP GUI. The market demand from SAP customers was very strong. 

The future of Quality Engineering: new roles and C-level attention

Does this mean Quality Engineers will eventually be replaced by AI?

Roman Zednik: No, I don't see that. But the roles will change. We've already gone through one transition from manual testers towards automation engineers. Now I think we will see people moving further towards roles such as quality architects and quality strategists.

They will use agents and steer agents, but they still have to validate the outcome. Is this sufficient? Is the result correct? Is it compliant? Is the system resilient? Are there issues around bias? These are questions Quality Engineering will increasingly have to address. We will see new roles around QA compliance as organisations begin to apply AI in regulated and business-critical environments.

You also predict that Quality Engineering could become a C-level issue. Why?

Roman Zednik: Look at what happened with cybersecurity. Fifteen years ago, security was largely regarded as an IT topic. Then major breaches happened, data was stolen and suddenly cybersecurity became a CEO and board-level concern.

I think something similar could happen with Quality Engineering. Once agents are operating inside business-critical processes, there will eventually be cases where an agent does something seriously wrong because it wasn't sufficiently quality assured. When that happens, quality will very quickly get management attention. I believe its importance will move to the C-level. 

TTC Global: So perhaps we'll eventually see a CQAO, a Chief Quality Assurance Officer, alongside the CISO?

Roman Zednik: Maybe something like that. I'm not sure what the title will be, but I do think Quality Assurance will become a C-level topic.

What does this shift mean for Quality Engineering partners such as TTC Global?

Roman Zednik: I see a very positive future for partners that adapt. Projects won't simply need manual testers or automation engineers. They will increasingly need quality strategists, risk architects and specialists who understand areas such as AI compliance.

That's where the methodology becomes very important. Risk-based testing is a good example. The idea that you shouldn't simply execute thousands of tests but should focus testing on the areas of greatest business risk has been around for a long time. In an AI environment, that thinking becomes even more relevant.

For partners like TTC Global, there is a bright future if they take their long experience in Quality Assurance and apply it to these new requirements around AI agents and testing. 

Finally, what advice would you give the CIO of an SAP organisation that is developing its AI strategy?

Roman Zednik: Every company needs an AI strategy. Ask what you actually want to achieve with AI and where it genuinely helps. And don't forget about the cost because that is becoming an important new dimension. But start early. There is a learning curve. People need time to get used to the technology and understand how to apply it.

I would create a small team focused on the AI strategy and ask: what can we do, and what can we realistically expect? Quality Assurance needs to be a core part of that strategy from the beginning.