AI is fast-moving, but so is our commitment to innovation. SAP has a long history of embedding AI capabilities across its portfolio.
Over the past few years, we have continued to lead the way with groundbreaking advanced AI capabilities for the enterprise.
This year at SAP Sapphire, we announced the latest innovations and highlighted the enormous momentum we are seeing with our customers and partners. We have delivered more than 230 AI scenarios.
By the end of 2025, SAP Business AI will have more than 400 AI scenarios on top of its 1,600 Joule skills, delivering unparalleled business value to customers. This commitment has already led to 34,000 customers utilizing SAP Business AI.
Innovation is at the heart of this growth, which is why SAP continues to enhance our generative AI copilot Joule, Joule Agents, and our AI scenarios embedded across the portfolio — all built with AI Foundation on SAP Business Technology Platform (SAP BTP), the technical backbone that powers Joule and Joule Agents.
We understand the importance of getting Joule and Joule Agents in the hands of our customers, especially in a competitive landscape where research predicts that a fifth of current sales team functions could be automated or that by 2028, 30% of Fortune 500 companies will only offer service through one AI-enabled channel.
During SAP Sapphire, we announced the latest SAP Business AI innovations that make Joule more powerful and accessible. I wanted to share a few innovations below, but I encourage you to dive into all the incredible announcements SAP made in the 2025 Innovation Guide.
Joule: the new UI in the age of AI
Our UI has continuously evolved to meet the needs of SAP customers. Now, SAP Business AI with Joule and natural language redefines the end user’s experience.
Joule is solidifying its role as SAP’s new UI, transforming how customers work and interact with our software through natural language — whether on the web, mobile, or in different languages. This way, the end user can ask questions in natural language, and Joule does the hard work in the system to find information or answer.
Joule is already one click away in every SAP application, but we also understand that customers store data across non-SAP systems, which should not prevent it from helping users. Joule cannot operate in a vacuum if it is going to offer real business value. That is why we are now expanding Joule to any application so that customers can access everything, everywhere, with the help of WalkMe.
Joule is breaking free to become omnipresent and accessible across non-SAP applications with the SAP Joule action bar supercharged by WalkMe. This integration expands Joule’s reach beyond SAP Business Suite, so it is always right by our customers’ side, giving Joule the “eyes, legs, and arms” to see what customers see.
Joule can follow business users across SAP and third-party systems, using context from active applications to connect data, surface insights, and streamline workflows.
No matter where our customers are, if they see information relevant to their business and want to know what it means in the context of SAP data, all they need to do is click to get the right context from Joule. This overarching context across multiple applications means Joule can orchestrate workflows and data from any application and website.
Joule is always on and proactive, providing customers with recommendations and proactively letting them know when Joule Agents have completed a task or require feedback. Customers are also empowered to create custom Joule skills for any application. Joule skills can handle repetitive, rule-based tasks, alleviating routine activities like data entry, IT ticket raising, and PO creation. Custom skills can execute these tasks based on natural language conversations with users that are triggered contextually and through user intent. This simple access through Joule only further democratizes access to AI and SAP.
Answer engine with Perplexity
If Joule is everywhere, it should also be capable of answering everything. Joule already provides answers across our customers’ structured and unstructured data, from procurement or travel activities to policies.
At SAP Sapphire, we elevated Joule again with help from SAP Business Data Cloud and SAP Knowledge Graph, and a new collaboration with Perplexity, an industry-leading company in the field of unstructured data.
SAP and Perplexity, an AI-powered answer engine that provides accurate, trusted, and real-time answers to any question, are collaborating to extend SAP’s generative AI and search capabilities by bringing together the best of both worlds in unstructured and structured data. By integrating Perplexity’s ability to address questions accurately and transparently with SAP’s business answer capabilities, we are bringing both worlds together: the power of Joule to answer business questions, and the power of Perplexity to understand what’s happening in the world.
Together with SAP Business Data Cloud, SAP Knowledge Graph, and our integration with Perplexity, Joule can now seamlessly combine internal business data with real-world insights. Whether data is structured or unstructured, from systems or the broader web, Joule brings it all together, delivering intelligent answers in real time. Learn more about the collaboration here.
Joule Agents: reason and act
Insight to action is a key principle of enterprise software. Businesses can access incredible insights by analyzing data, which they can use to arrive at conclusions before making a decision. Joule makes getting these insights through natural language simple. Now, Joule Agents can reason to fulfill the action part of the equation with other agents. This ability marks a paradigm shift in enterprise software that helps customers go from insight to action to reason and act with AI.
Joule Agents can reason and coordinate to autonomously execute tasks through multi-step, self-conceived workflows while always involving a human.
Envision an environmental health and safety agent assessing a new chemical entering a plant: This agent could ensure that inventory is not stored near reactive chemicals and then could consult with HR to verify that staff are qualified before recommending appropriate protective equipment. Additionally, the all-new accruals agent for finance reduces today’s manual work by 90 percent. This agent learns a customer’s accounting rules and then, using financial data, can propose the right accruals during financial close.
In this sense, as much as Joule is the new UI for AI, Joule Agents are reimagining the business process layer across SAP Business Suite in an AI-native way.
AI agent hub in SAP LeanIX
As the number of agents across SAP and third-party vendors grows, customers will need a way to discover agents and then govern and control what they do. This is precisely the role of the AI agent hub in SAP LeanIX solutions: It offers a governance and control framework to manage all these agents.
Because SAP LeanIX already knows our customers’ enterprise landscape, it can map every available AI agent to the business capability or process step it supports. This framework helps customers spot automation gaps and discover new AI agents to fill them. Tracking governance for each agent means customers can, for example, make sure agents meet ethical and compliance standards.
The AI agent hub in SAP LeanIX will be generally available in the fourth quarter of 2025.
SAP, NEURA Robotics, and NVIDIA
SAP announced a collaboration with NEURA Robotics and NVIDIA, unveiling a new era where SAP Business AI moves to the physical world.
Through this collaboration, Joule Agents connect enterprise data and processes with NEURA’s advanced cognitive robots, ranging from mobile autonomous mobile robots (AMRs) to humanoids, empowering businesses to automate real-world tasks such as adaptive manufacturing, autonomous replenishment, compliance monitoring, and predictive maintenance. The Mega NVIDIA Omniverse Blueprint plays a crucial role, enabling enterprises to validate and optimize robotic fleets in digital twins before deploying them on the shop floor or in the warehouse, thus ensuring safety, efficiency, and rapid scaling.
AI Foundation: the AI operating system business AI
We knew from the outset that Joule and Joule Agents needed a strong operating system. That is what SAP’s AI Foundation on SAP BTP, our operating system for AI, now offers customers. AI Foundation gives developers a centralized set of tools to build, extend, and run custom AI solutions and agents at scale, all accessible via one entry point. AI Foundation already had more than 90 capabilities, and at SAP Sapphire, we introduced over 60 more, such as AI Agent runtime, a brand-new AI evaluation service.
AI Foundation accelerates AI development by bringing together SAP Business AI solutions such as generative AI hub, Joule Studio, SAP Document AI, and SAP Knowledge Graph. Centralizing and simplifying AI operations also reduces the total cost of ownership.
SAP Knowledge Graph gives Joule and Joule Agents more accuracy and context. It grounds AI in a customer’s business context and exposes metadata that makes the relationships between information accessible. This way, for example, an agent working on a customer service ticket knows the relationships between products and purchase orders. Initial support for the SAP Cloud ERP and SAP Business Data Cloud solutions is expected to be generally available by the third quarter of 2025.
Other SAP Business AI solutions are also getting new enhancements. For example, Joule Studio, which helps developers create custom AI, will get a new skill builder capability to help developers create custom skills for Joule. We are also adding an agent builder capability that allows developers to create custom AI agents using low-code and no-code capabilities. General availability is planned for the third quarter and fourth quarter of 2025, respectively.
AI Foundation offers the latest AI models, including GPT-4.1, Claude 3.7 Sonnet, and, most recently, Gemini 2.5 Pro. Customers can select from a range of self-hosted models on SAP’s GPU infrastructure, such as Mistral AI, Cohere, or Meta model families for more sovereignty and control. Customers can also pick from best-in-class, industry-specific models created by our partners. One example is Deloitte, which brings its unique finance and tax experience to the AI Foundation with Zora AI.
Deloitte plans to integrate Zora AI with SAP Business AI, enabling customers to create intelligent digital workforces powered by agentic AI. Using SAP BTP, Zora AI offers a suite of preconfigured agents capable of autonomously performing complex business tasks with precision and speed. When paired with SAP’s robust enterprise capabilities, and enhanced by the data access in SAP Business Data Cloud, Zora AI has the potential to transform how work gets done, helping customers unlock the full value of agentic AI. Learn more about the integration here.
All these models deliver incredible results when the prompts are carefully designed. But herein lies a significant challenge: Once our customers are committed to a specific model, it can be difficult to change, as even subtle model updates can break carefully tuned prompts. That is why we worked hard to eliminate prompt engineering.
Prompt engineering is dead
Prompt engineering is dead; say hi to benchmark engineering. Prompt engineering requires customers to learn how to “talk” to the AI so it can effectively leverage its knowledge and capabilities to meet their specific needs. This process was manual and cost-intensive; above all else, it made it tedious and slow to change models.
I am excited to introduce SAP’s prompt optimizer, designed collaboratively with Not Diamond, a frontier AI lab renowned for its model-agnostic approach. Prompt optimizer is a capability of the generative AI hub in AI Foundation.
Switching models and optimizing prompts is time-consuming — it can take an organization days to migrate a single use case from one model to another — but it shouldn’t be. We wanted to tackle that challenge because customers need to be able to change models easily. Maybe a new model is better, more cost-effective, or unavailable in your country: Whatever the reason, switching models shouldn’t be difficult.
SAP’s prompt optimizer automates this entire task. Selecting a new model automatically converts all existing prompts into optimized prompts for all the latest models using public benchmarks. This helps reduce the time needed to change models and optimizes them from days to only minutes of background processing.
The prompt optimizer now means SAP customers can switch to the best available AI models for their specific use cases when they become available without any costly delays. It lays the groundwork for a multi-model future, where applications are decoupled from specific models and customers can utilize the most suitable model anytime.
We want organizations to seamlessly use AI capabilities without being locked into single-model dependencies, fostering flexibility and innovation in building AI solutions. A limited early access program is now available for early sign-up, with general availability expected later this year.
SAP Document AI
These announcements, and many others, made at SAP Sapphire in 2025 reaffirm our commitment to continually innovate and listen to feedback to meet our customers’ evolving needs. Take paperwork: Billions of documents are scanned into SAP systems yearly. Extracting relevant information from documents manually is costly, error-prone, and inefficient. That is why I am excited to introduce SAP Document AI, previously Document Information Extraction, which is natively integrated with SAP applications.
SAP Document AI seamlessly processes, extracts, and understands structured and unstructured data from business documents while automating document processing. SAP Document AI significantly reduces manual efforts and minimizes errors. Customers like Serax, a Belgian design brand, already use SAP Document AI to reduce manual work by 33 percent.
SAP Document AI will build upon the well-established Document Information Extraction Base and Premium editions, preserving their proven capabilities while enhancing efficiency. In addition, later in 2025, we plan to introduce the Premium Plus edition to expand the product’s scope, ensuring smooth, out-of-the-box integration across the SAP portfolio. For more details, read “SAP Document AI: The Road Map for Seamless Document Processing in SAP Business Suite.”
We continually strive to innovate our products; the same applies to SAP Document AI. That’s why we introduced instant learning to enhance the out-of-the-box accuracy of SAP Document AI in the pursuit of reaching 100 percent. Suppose SAP Document AI extracts something incorrectly from a document due to improper formatting. In that case, customers simply need to log in and correct the error once for the system to learn and avoid making the same mistake again. Previously, this would have required time-consuming retraining with numerous documents. To see it in action in SAP Document AI and SAP Cloud ERP, watch this demo video.
Additional AI innovations with partners
- Cohere: SAP is expanding its partnership with Cohere to bring secure, enterprise-grade generative AI models to the SAP ecosystem, starting with Cohere’s Rerank model and extending evaluations with the Command and Embed models. These models are planned to be available alongside other leading AI models from SAP and third parties in the generative AI hub in SAP AI Core, with the intent to give customers more choice to build AI-powered solutions that meet their unique business needs.
- Vianai Systems, Inc.: Vianai’s hila is an innovative generative AI conversational analytics application designed to engage a series of line-of-business executives. It currently focuses on empowering the office of the CFO, enabling direct interrogation of transactional systems. Our collaboration with Vianai centers on exploring how hila could complement established platforms such as Joule and the just ask feature in SAP Business Data Cloud by offering greater query precision and accuracy for non-SAP systems in the future. We are energized by the opportunity to push the boundaries of conversational analytics together. Stay tuned as we continue to uncover the transformative potential of these advancements. Learn more about the collaboration here.
- AWS: SAP and AWS have announced the launch of an AI co-innovation program to help partners create generative applications, providing customers with direct access to innovative solutions that address complex business, logistics, and supply chain challenges. The program includes a fund in the form of AWS and SAP technical resources, expertise, and credits to help customers and partners leverage Amazon Bedrock through SAP BTP to define, test, and deploy industry-specific generative AI solutions, such as forecasting and planning for severe weather impacts on operations or increasing precision in financial outlooks. The program underscores the companies’ shared vision to help customers develop and deploy practical, secure generative AI applications that solve critical business operations challenges. Learn more about the co-innovation here.
AI investments
The externally managed SAP.iO Fund has made an investment in Nova Intelligence. Nova Intelligence is building end-to-end AI coding agents that can parse and understand legacy custom ABAP SAP code to modernize it into cloud-compliant applications. The AI agents iteratively test and refine the generated code until a successful outcome is achieved. Nova Intelligence is developing specialized AI agents to support the transformation of legacy custom code. Learn more about the investment here.
New SAP Business AI capabilities as part of the SAP Sustainability portfolio
We’re introducing new AI capabilities in the SAP Sustainability portfolio to help customers prepare data, automate processes, and enhance expertise. These innovations streamline environment, health, and safety (EHS) and product compliance processes and optimize risk and cost management while advancing sustainability goals. They will be available in beta in August 2025. New capabilities include:
- AI-assisted permit management in SAP S/4HANA for EHS: Automates and simplifies permit handling, extracts key data, and proposes follow-up tasks
- AI-assisted safety observation reporting with Joule in SAP S/4HANA for EHS: Enables seamless incident reporting through a conversational interface
- AI-assisted safety instruction generation in SAP S/4HANA for EHS: Suggests safety instructions based on risk assessments
- AI-assisted compliance information processing in SAP S/4HANA for product compliance: Automates data extraction and mapping from declarations and certificates
These capabilities will be available in the new Joule Premium for Supply Chain Management package, adding to existing AI features like:
- AI-assisted ESG report generation in SAP Sustainability Control Tower
- AI-assisted emission factor mapping in SAP Sustainability Footprint Management
- AI-assisted declaration image analysis for SAP Green Token
Combining existing and new AI features, SAP Business AI helps customers reduce costs, lower risks, and drive performance in reporting, carbon management, circularity, and compliance.
SAP Business AI in 2025
SAP Sapphire was an exciting time. We have announced many innovations that drive value for our customers; however, the rest of 2025 promises to deliver more. We will continue offering AI scenarios that help our customers, from new Joule Agents to building strong, open partnerships that realize value in this fast-paced AI era.
To learn more about how SAP Business AI delivers innovations, watch the SAP Sapphire keynote, read the 2025 Innovation Guide, and visit the SAP Business AI area of sap.com.
Philipp Herzig is CTO and chief AI officer and a member of the Extended Board of SAP SE.
