Download the full interview here.
SAP programs are being reshaped by three transformative forces: the clean-core mandate, an agentic approach to automation, and a heightened focus on developer impact. The result isn’t merely a new slogan, but a fundamentally new working model where architecture, security, and data movement are designed as an integrated whole.
Soulat Khan leads SAP Solution Architects worldwide at Amazon Web Services, where his team develops partner solutions and AWS services for SAP customers. This includes publishing joint reference architectures that provide guidance on adopting SAP Business Technology Platform and AWS agentic services, such as Amazon Bedrock and supported Large Language Models (LLMs). Through co-innovation with partners and customers, he maintains a close view of where AI is already transforming day-to-day SAP work, how teams are leveraging Amazon Q Developer to interpret and explain legacy ABAP code, and what near-term progress is realistic versus hype.
In this interview with ASUG, Khan discusses the rising influence of developers within the enterprise, what “resilience” means when cloud placement and data movement choices affect outcomes, and why secure interoperability matters as agentic AI becomes more prominent.
This interview has been edited and condensed for length and clarity.
The Agentic AI Revolution
Q: Regarding agentic AI, what have you observed as the most game-changing moments for customers in the past six months, and what excites you most about the trajectory you’ve been laying out?
The most exciting development is that customers can now build entirely new business processes by enabling data flow across their enterprise systems.They no longer need to operate within particular silos or build custom integrations to access, manage, or utilize their data for creating new experiences and applications. Agents can handle that communication on their behalf.
The adoption of frameworks like Model Context Protocol (MCP) and Agent2Agent (A2A) is facilitating the establishment of security, governance, and interoperability. We’re proud to be part of that partnership with companies like Anthropic and SAP as part of our expanded independent software vendor (ISV) portfolio. By enabling bidirectional data communication with SAP, customers gain access to combine their data assets in ways they previously couldn’t.
Many companies have invested heavily in customizing access to ERP and non-ERP systems. Now, as we progress into this agentic phase, we’re building models and agents to handle that communication on our behalf, which is exciting.
The era of the ABAP developer focused solely on driving Z codes, transaction codes, and unit testing multiple entries will evolve into full automation of business suites, integrating services from SAP, Amazon, and any other existing applications that now interface with an agent to build that communication. We’ve already begun this work through our AI innovation program for SAP with partners like Accenture, Deloitte, Capgemini, and many others.
The AWS AI Co-innovation program with SAP is where we work with partners not only to enable SAP integration, but to build agentic solutions using AWS and SAP services. Customers can obtain these solutions out of the box, purchase them from a marketplace, implement them in their current SAP environment, and immediately begin realizing value.
Defining Resilience
Q: You frequently discuss resilience, which underpins much of what you do at AWS. You’re also the author of “Building a Resilient SAP Enterprise.” What does resilience mean to you at a high level, how do you measure it, and how does it help move AI from theoretical to operational?
If SAP is the heart of your business, then choosing the cloud is like selecting the arteries through which all vital nutrients are provided to the body. Just as you don’t want your arteries to fail, you don’t want your cloud provider to stop serving you. Helping customers realize they have choices in their enterprise architecture makes them more resilient. It’s not simply “buy this suite of applications and deploy them wherever you want.”
You can evaluate each set of applications, in addition to the core platform, and make precise decisions to build resilience for your specific business scenarios and workflows. At Amazon, we’re customer obsessed. First and foremost, we want to ensure our customers are thinking about how to deliver value in their business when they deploy an SAP workload. What’s the best adjacent workload to help drive outcomes for them as well?
For example, when deploying SAP systems, where and how you deploy them matters. If you’re deploying SAP Business Technology Platform (BTP) applications, where those applications reside can matter, and how they interact with your internal AWS systems and RISE systems can also matter.
Elevating Developer Influence
Q: Considering developers’ role in business, and in the context of the agentic evolution, how are you working with developers to elevate their influence? When they want the attention of business leaders and need to build the case for these developments, what do you tell them?
The cost of entry for development has dramatically decreased. You no longer need to purchase expensive systems or large software licenses. You can work in an agentic integrated development environment (IDE) and open a free trial to process your legacy code and drive real results. I encourage experimenting, exploring, and learning about these new services.
Even before building, evaluate whether you truly have a problem to solve. As engineers and ABAP developers, we can help identify problems before they become business disruptions and create new agentic solutions quickly using services and tools. At AWS, we’re delivering primitives and tools to help customers build next-generation experiences for their business and clients. My team hosts monthly enablement workshops to explain and showcase how customers and partners can leverage these tools to use them in their environments — I encourage folks to attend those.
Security and AI Collaboration
Q: Much of that exploration involves the SAP Discovery Center and published reference architectures. You also mentioned developers collaborating with security teams so everyone understands what AI means for data interaction, models, and production. How has that relationship evolved, and why is it important as we explore the agentic path?
As we always say at AWS, “security is job zero,” and this has never been more critical than in today’s rapidly evolving agentic landscape.
Let me put this in perspective: in a typical enterprise environment, an employee might interact with at least several different applications daily. Now, when we enable an AI agent to perform similar tasks, it needs to navigate these same systems but with added complexity. While traditional API management has given us a strong foundation for secure interactions, we’re entering uncharted territory when we allow LLMs to autonomously decide which tools to invoke.
Connect with Soulat Khan on LinkedIn.
The challenge becomes even more intricate when we consider Agent-to-Agent (A2A) interactions. We’re not just dealing with traditional authentication and authorization flows anymore; we’re managing AI agents with billions of parameters and their own knowledge bases interacting with each other. This creates new security considerations around data access patterns, permission boundaries, and interaction monitoring.
This is precisely where Amazon Bedrock plays a crucial role. It provides a fully managed service that gives enterprises access to leading foundation models through a single API, all while keeping your data and prompts private and secure. This is critical because it means you’re not sharing sensitive enterprise data with public endpoints. Everything stays within your AWS security boundary.
Bedrock’s integration with AWS Identity and Access Management (IAM) provides granular control over which teams and applications can access specific models and capabilities. This is particularly important in the agentic landscape we discussed, where you need precise governance over AI interactions. You can define exactly which agents can invoke which models and under what conditions.
Furthermore, Amazon Bedrock AgentCore Identity and AgentCore Gateway are specifically designed to address these emerging challenges. These services help manage agent identities, control access patterns, and provide secure gateway capabilities for agent interactions, ensuring that all AI-driven operations maintain enterprise-grade security standards.
Amazon Q Developer and ABAP
Q: Let’s dive deeper into Amazon Q Developer, especially its ability to explain business processes in ABAP. At last year’s SAP Sapphire & ASUG Annual Conference, NVIDIA CEO Jensen Huang described ABAP as “the most valuable language at SAP.” What does it mean to interpret, read, create, and identify connections in ABAP, and what does that make possible for developers?
You no longer need to ask around about what code means or experiment to understand what code does. You can run Amazon Q in your code through the Eclipse ADT plugin and immediately understand what your application is doing. It can understand and generate test cases, help with unit testing ABAP code, and generate ABAP code for your requests that meets your company’s standards. You can ask it to integrate with AWS services where appropriate.
We are leaning in to partner with SAP on RISE. As customers move to RISE, they need to remediate legacy code, which was previously a weeks or months-long process. With Q Developer, customers can reduce this effort by 90% or more, de-risking and reducing the cost of going from their legacy, ECC-based processes to modern processes that run in SAP S/4HANA Cloud and BTP.
Learning and Upskilling Platform
Q: You also position Q Developer as a learning tool and platform. What feedback have you seen from customers, like Zappos, about why it’s effective for upskilling across complex, multi-instance SAP landscapes?
Many of our large enterprise customers don’t have a single SAP footprint. They have multiple instances, sometimes across multiple regions or versions. Standardizing their code base is challenging. Sometimes the original developers are no longer with the company, or the code is written in a way that’s difficult to understand. You can use Amazon Q Developer from an attached chat window inside your Eclipse IDE to interpret code in natural language.
See Amazon Q Developer in action.
If you want to customize it for your business needs quickly, you can do that, too. Ultimately, however, we want you to minimize customizations and modernize for a clean core — remove unnecessary code and leverage services, whether from BTP or AWS services directly. That’s one of the most powerful capabilities we provide.
Q: What else about Q Developer would you like to highlight?
For customers interested in building agents and becoming an agentic enterprise, understanding what your code does today will help advance that journey. Testing and exploring with Amazon Q Developer will give you better understanding of your code base. As you progress into creating agents, having a solid foundation of what the code base is doing will lead to better success in your agentic workflows.
Looking Ahead to 2026
Q: Looking ahead to 2026, what prediction would you make for what we’ll be discussing this time next year?
Consider this: Adidas is one of our SAP customers. They leverage Bedrock to build brand-new shoe designs. They’ve trained over 150,000 shoe designs in Bedrock and can use natural language to generate new inspired shoe designs with influences like Spanish artist Gaudí. They’ve also implemented Buy with Prime directly on their website, so customers familiar with Amazon’s buying experience can visit Adidas, purchase shoes and clothing, and use Prime shipping as part of the delivery process.
This time next year, we will see this be the norm for customers who choose AWS for SAP — not just running their SAP applications on the most secure, reliable, and extensive cloud infrastructure, but going on a journey with AWS and Amazon as their business transformation partner. By tapping into our global scale, operational efficiency, supply chain, and fulfillment operations, customers can transform their core business processes. This is including everything from warehousing, fulfillment, advertising, and e-commerce to reimagine the entire customer experience end to end.
Advanced Agentic Capabilities
Q: When you think about a less obvious agentic capability, is there one you would point to as organizations build agents that can do things people might not typically expect?
Let me share an interesting example that showcases the less obvious but powerful potential of agentic capabilities. Amazon Rufus, which we recently launched for Amazon Shopping, demonstrates how AI agents can transform seemingly straightforward experiences in unexpected ways. While shoppers might traditionally search for products using keywords or categories, Rufus can engage in natural conversations about products, understanding nuanced contexts like “what makes a good running shoe for beginners” or “how to choose between different types of coffee makers.”
What’s particularly fascinating is how Rufus combines product knowledge with contextual understanding. It’s not just matching keywords; it’s understanding the intent behind questions, considering factors like user experience level, use cases, and even potential concerns that the customer might not have thought to ask about.
This points to a broader opportunity in enterprise contexts. Imagine similar capabilities in SAP procurement systems, where agents could help procurement specialists not just find products, but understand complex supply chain implications, evaluate sustainability metrics, or navigate regulatory requirements across different regions — all through natural conversation.
Visit the AWS website.
The key insight here is that agentic capabilities aren’t just about automating existing processes; they’re about unlocking new ways of interacting with and understanding complex business systems. We’re seeing organizations build agents that can interpret and act on unstructured data, understand context across multiple domains, and provide insights that might not be obvious even to experienced professionals.
Agents that can spawn new agents to interact with multiple data systems aren’t something you typically consider. If you need to curate system support data that you don’t even know you need, you can ask the agent to think about that and build on it in a way that abstracts those data systems. We do that with Strands. You can design agents to communicate with other agents or spawn new agents.
The delivery and deployment can be anywhere in the enterprise — in SAP systems, in AWS systems, really anywhere you choose for an agentic enterprise. If you don’t know where to start, we can match you with a partner or build an organizational change roadmap to determine the best landing spot for those agents.
Q: Can you share an example where that capability has been revolutionary for a business?
I’d encourage you to join us for re:Invent where we’ll showcase several new innovative examples with partners like Hyundai, Mondelez, and Colgate, and make some new compelling announcements. At re:Invent we’ll cover the entire SAP journey—from migrating the SAP applications you have to AWS to improve security, agility, and reliability—to moving to S/4HANA, unlocking the value of your data, and transforming your operations with the power of AI.