SAP: AI in 2026 – Five defining topics

To steer the development of AI, companies need to understand that it’s no longer just a question of “What can AI do?”, but “How do we get our company on the road to success with AI? What do we need to do? Which problems do I solve with which models? How do we manage all of this?”

Let’s take a look at five important topics that will define corporate AI in 2026 and that represent both opportunities and challenges for companies. Here we go!

1. New categories of AI base models create added value for companies

The advances in generative AI are due to breakthroughs in so-called base models. These are large-scale neural networks that are trained with enormous amounts of data and can be adapted to a wide variety of tasks.

Large language models (LLMs) were the first wave of comprehensive base models. General LLMs trained with the equivalent of all text on the Internet enabled many value-added use cases such as summarizing documents, writing code, and supporting applications like ChatGPT and Claude. In recent years, we have already seen the base model approach applied in other domains, such as video creation and speech generation.

2026 Specialized base models optimized for specific data types and domains will support high-value use cases for enterprise AI. In video generation models, it has already been observed that models based on real physical data can draw conclusions from scenes and physical dynamics. Emerging world models show that simulating the physical world opens up new possibilities in the areas of simulation, synthetic training data and digital twins. Vision-language-action models demonstrate that robotic base models can derive generalizations and apply them to new tasks and environments, enabling the transformation of web-scale knowledge into real-world actions in logistics and manufacturing.

In the enterprise space, a similar transformation is taking place with respect to structured data in databases and transactional enterprise software. Although LLMs deliver impressive results in many enterprise use cases, they cannot solve tasks like numerical forecasting, such as deriving a delivery date or supplier risk assessment. However, work on relational base models shows that training with structured data sets – for example, data in tables rather than general text or images from the internet – enables high accuracy predictions without the tedious feature engineering and training required by traditional machine learning. This means that companies can implement predictive models in a matter of days rather than months. Recently introduced relational base models such as SAP-RPT-1, Kumo and Distil Labs demonstrate how new models can directly support use cases such as forecasting, anomaly detection and optimization in ERP, finance, manufacturing and supply chain scenarios.

These specialized models are expected to scale in 2026 and deliver best-in-class, efficient performance for structured business tasks. They will outperform general LLMs and modern machine learning algorithms. These models will be the future drivers for value-adding tasks in organizations.

2. Software evolves towards an AI-based architecture

The field of AI has seen a variety of value-adding approaches over the decades – from the first rule-based expert systems to probabilistic deep learning and the recent boom in generative AI. In 2026, organizations will move from augmenting existing AI applications and processes to AI-based architectures that will fully deliver on the promise of modern AI.

AI-based architectures add a layer of continuous learning and agentic AI to deterministic systems. This enables intent-driven and context-aware applications that are self-improving and do not need to be statically coded around fixed workflows. However, agentic systems will continue to be only as good as the context layer on which they are based and from which they can reliably retrieve information. With this in mind, organizations should invest in comprehensive, semantically rich knowledge graphs that provide a scalable source of context and enable AI-based software that is reliable and self-improving.

Enterprise applications will increasingly build natively on AI capabilities and provide user experiences designed for interactions with multiple natural language models. They will be equipped with AI agents that can reason across complex processes and have a foundation to manage base models, services and a knowledge graph to capture semantically rich business data. An AI-based architecture will also enable more employees to quickly create apps themselves – for example, smaller ad-hoc productivity applications – without the need for IT teams.

A prerequisite for an AI-based architecture is established SaaS principles and investment in modern cloud applications on which it can build. The technical term for the combination of probabilistic, adaptive AI models with deterministic recording systems is neurosymbolic AI. It combines the best capabilities of AI to adapt to reliable, controllable and deterministic processes. Next-generation applications will not only support AI, they will be built on AI at their core. This means combining reasoning, business rules and data to seamlessly deliver and automate insights. Imagine ERP systems that proactively report anomalies, recommend actions and even execute workflows autonomously – all in line with corporate policies and regulatory requirements.

3. Agentic governance becomes a business-critical aspect

In the last two to three years, generative AI has triggered a wave of value-adding use cases. These use cases were largely based on the following pattern: users send a prompt to a model, receive a response and interact with the model again.

Last year, the next wave of innovation rolled in: AI agents that can plan and iteratively reason through multi-step tasks, including selecting tools, self-reflecting on progress and collaborating with other AI agents. These advanced AI agents promise to tackle complex business processes that previously could not be automated, such as analyzing a large number of documents, records and policies to resolve a dispute or book a trip.

However, the proliferation of AI agents, many of which process critical tasks and sensitive data, will require the development of new capabilities. Agentic governance will emerge as an important function as companies deploy hundreds of specialized AI agents. The challenge of “agent proliferation” will be reminiscent of previous shadow IT crises, but it will come with higher risks given the autonomous decision-making capabilities of agents.

Future-oriented organizations will create comprehensive governance frameworks that cover five dimensions: Agent lifecycle management (version control, test logs, deployment approvals, retirement procedures), observability and auditability (agent directory, logging, reasoning paths and action traces), policy enforcement (embedding business rules, legal constraints and ethical guidelines into agent execution workflows), human-agent collaboration models (defining autonomy boundaries, approval requirements and escalation paths) and performance monitoring (tracking accuracy, efficiency, cost and business impact).

The organizational transformation will be profound – from viewing AI as an independent tool to managing agents as digital employees requiring onboarding, performance reviews and continuous improvement. HR and IT departments will collaborate on “digital workforce management” as organizations take agentic governance as seriously as traditional workforce oversight.

4. Intent-driven ERP and generative UI offer new user experiences

Consumers are increasingly comfortable with computer interactions that require prompts in natural language, voice input and even images and gestures. At the same time, the ability of generative AI to dynamically create text, graphics, code and HTML will improve rapidly. In parallel, it will be possible for users to simply express their intentions so that AI agents can determine how to achieve this goal.

These advances will open up a variety of completely new ways for users to work with enterprise software and with ERP software without an app (“no-app ERP”). Let’s take the example of booking a customer visit. To do this, employees usually have to open an analytics application to check the customer account. Then they search for the customer’s address in the CRM system and finally navigate to another application to book the trip. 

2026 We will work with digital assistants more often via “Gen UIs”. Generative UI relieves users as they no longer have to navigate between multiple applications and perform manual tasks. Over time, AI will allow users to simply express their intent: “Prepare a trip to my customer with the most leads.” The AI agent then plans the appropriate steps and required systems and interacts with the user to confirm trip details. Meanwhile, it generates dynamic analytics charts and briefing material in the window. As AI agents develop more powerful computational and predictive tools, they will allow users to “speak more naturally to their data” while the agents make data-based decisions in the background. To clarify once again: Interactions with agents will go far beyond common chat dialogs. Companies will have extensive visualizations, complete workflows and the ability to create highly personalized apps with just a few commands.

The user interface itself will not disappear. Working with ERP independently of an app and with autonomous agents requires the same fundamentals that people already rely on in their daily work: structured workflows, security, governance and business logic defined in business applications. The difference is that agents consume these basic operations programmatically at scale, not just through a graphical user interface. Humans can interact with these agents via natural language without even having to open the application.

These capabilities will establish a new paradigm for human-AI collaboration and workplace productivity. Personalized work and adaptable workflows across applications and data sources will improve adoption. This ability to focus solely on achieving a user’s intent, regardless of interaction modality and underlying systems, will increase the return on investment (ROI) of AI and enterprise software.

5. Deglobalization increases the supply of sovereign AI

AI has created debates about digital sovereignty between different countries as it potentially impacts everything from scientific discovery to national security to economic productivity and even culture. Geopolitical events such as supply chain disruptions from tariffs and war have added to the urgency of achieving digital sovereignty felt by many nations and businesses.

Digital sovereignty has two broad definitions. First, digital sovereignty is a term for information security that regulates the storage and access of data, such as FedRAMP in the US and the VSA in Germany, which are prerequisites for processing sensitive government data in a sovereign cloud. In the second, more general definition, sovereignty refers to the origin of physical assets, intellectual property, jurisdiction and services for the entire cloud stack. For example, does an application use an AI model built in Europe, the US or China, and is the data center geographically isolated?

Given the high risks, geopolitical uncertainty and complexity of “sovereign AI”, companies will increasingly demand AI and cloud solutions that are simultaneously cutting-edge, flexible and fully sovereign. This is driving the shift from a globalized one-size-fits-all cloud to regionally compliant, AI-powered enterprise platforms. At the same time, governments will continue to refine their national AI strategies to invest in areas along the stack where they can compete and add value.

Implementation of AI themes in 2026

By 2026, AI is well on its way to evolving from a supporting tool to a mainstay for businesses. This transformation is being driven by a convergence of key trends – including increasingly powerful agents, generative user interfaces and AI-based architecture – that are moving AI from the application layer into the core of business operations.

The companies that will succeed are those that recognize this shift and build an enterprise specifically focused on AI: creating robust governance to manage a new, collaborative workforce of humans and AI agents. They introduce generative user interfaces to improve adoption and enable users to work with intent through natural interaction. They are unearthing specialized base models that precisely match the organization’s use cases to drive business value, and last but not least, they are seamlessly developing applications around AI that unite logical reasoning, business rules and data to deliver proactive insights and enable automation.

However, organizations will continue to need high-quality, connected data in 2026. Data silos severely limit the effectiveness of AI. As mentioned earlier, an AI-based architecture requires investment in modern cloud applications that harmonize data across the enterprise – because unified data ensures that AI results are more accurate and relevant.

________________________________________

Jonathan von Rueden is Chief AI Officer at SAP SE.
Walter Sun is Senior Vice President and Global Head of AI for SAP Business AI at SAP.
Sean Kask is Vice President and Head of AI Strategy for SAP Business AI at SAP.

– – – – –

Further links

👉 www.sap.com 

Photo: pixabay