Let me start with a simple experiment: Ask a generative AI tool to count the words in a document. It will likely be off by 10%.

In a blog post, that’s tolerable. In a financial disclosure, a regulatory filing, or a supply chain commitment, it is simply unacceptable.

Generative AI is statistical. Answers to enterprise level problems are a lot more deterministic. The distance between 90% and 100% accuracy is not incremental. In our world, it is existential.

In 2026, AI is no longer evaluated on novelty. It is evaluated on precision, governance, scalability, and business impact. As organizations move from pilots to scaled programs, five moments will define whether they capture lasting value or expose themselves to avoidable risk. I have seen these moments play out across every major market I oversee.

1. The governance moment: when agents become digital coworkers

The first moment arrives when AI stops being a tool and starts being an actor.

Agentic AI systems plan, reason, orchestrate with other agents, and execute workflows autonomously. They touch sensitive data and influence decisions at scale. If you are not already governing them as you govern your human workforce, you are exposing your organization to risk.

Agent sprawl will mirror the shadow IT crises of the past decade, but the stakes are categorically higher. Enterprises must establish agent lifecycle management, clear autonomy boundaries, policy enforcement, and continuous performance monitoring. Every board needs to answer three questions: Who is accountable when an agent makes the wrong call? How are decisions audited? When does the machine escalate to a human?

Geopolitical fragmentation compounds this urgency. Sovereign cloud, sovereign AI, and data localization are no longer theoretical concerns. They are regulatory realities in markets from New York to Frankfurt to Riyadh to Singapore. Governance in the age of AI is less about controlling risk at the edge and more about embedding deterministic control into probabilistic intelligence. That is a C-suite mandate, not an IT project.

2. The data foundation moment: when the last mile is the only mile that matters

The second moment is quieter, but it is where most enterprises will ultimately win or lose.

AI is only as reliable as the data and processes it operates on. Fragmented master data, siloed systems, and over-customized ERP landscapes introduce unpredictability at the worst possible moment: when AI provides a recommendation that affects your customers, your cash flow, or your compliance position.

Enterprise AI value will not come from generic large language models trained on internet-scale text. It will come from intelligence grounded in your enterprise data—orders, invoices, supply chain records, financial postings—embedded directly in your processes. Relational foundation models optimized for structured business data will outperform generic LLMs in forecasting, anomaly detection, and operational optimization.

The question every board should be asking is not only “What AI can we add?”, but also, “Is our data estate ready, or are we layering probabilistic intelligence onto fragmented foundations?”

3. The employee interaction moment: when the interface disappears

The third moment happens in your employees’ daily workflows, and it will accelerate faster than most organizations expect.

In 2026, we are moving from static application interfaces to generative user interfaces. Instead of navigating between systems, employees express intent: “Prepare a briefing for my highest-revenue customer visit this week.” AI agents orchestrate the workflows, assemble the context, and surface recommended actions.

But adoption is not automatic, and trust is not given. Employees will embrace AI teammates only when they are confident that outputs respect governance boundaries, reflect real business rules, and deliver measurable gains. Role-specific AI personas tailored for the CFO, the CHRO, the head of supply chain, built on trusted data and embedded in familiar workflows, are what will close the adoption gap.

Organizations that invest in AI-native architecture will accelerate ROI. Those that bolt AI onto legacy interfaces will struggle with trust, usability, and scale. This is a design decision with strategic consequences.

4. The customer moment: when intelligence becomes a competitive moat

AI proves its enterprise value most visibly at the customer edge.

Trained on your own data, your own policies, and your own interaction history, customer-specific intelligence compounds in ways that competitors cannot easily replicate. This is especially powerful in exception-heavy environments: dispute resolution, claims handling, returns management, service routing. AI that can classify cases, surface relevant documentation, recommend policy-aligned resolutions, and learn continuously from outcomes transforms these high-cost, high-friction processes into sources of competitive differentiation.

In 2026, your customers will not reward novelty. They will reward reliability, relevance, and responsiveness. Organizations that use AI to absorb complexity, without losing control over outcomes, will build moats that generalist tools cannot breach.

5. The strategy moment: when you decide how far to go

The final moment is the one that falls squarely on leaders.

AI adoption is not a single journey. It requires leaders to orchestrate three layers in parallel:

• Embedded AI: Persona-driven productivity gains built into core applications for immediate returns
• Agentic AI: Multi-agent orchestration of complex, cross-system workflows
• Industry AI: Deeply specialized applications co-developed to address the highest-value challenges specific to your sector

The trap is false sequencing: focusing only on embedded AI leaves value on the table and jumping to deep industry transformation without governance and data maturity multiplies risk. The organizations that will lead are those that align ambition with readiness and invest in clean core architecture, modern data foundations, and cross-functional AI ownership, while moving decisively from pilots to programs.

The leadership test

In 2026, the winners will not be those with the most AI features. They will be those who treat AI as a core operating layer, governed like a workforce, grounded in trusted data, tailored to employees and customers, and calibrated to the realities of their industry.

The gap between 90% and 100% is precisely where enterprise value lives. It is also where leadership is tested. The decisions you make in the coming months will determine whether AI becomes your most powerful source of durable advantage or your most expensive lesson in misplaced confidence.

This is the moment to move with precision.

Manos Raptopoulos is global president of Customer Success Europe, APAC, Middle East & Africa, and a member of the Extended Board SAP SE.

The landscape of user interfaces is undergoing a seismic shift. The explosion of consumer AI has reset expectations for business software: Employees now expect their enterprise apps to have the same intuitive, conversational interfaces they use at home.

This has led to a “Terminal Renaissance,” a return to text-in, text-out interaction.

For many applications, text works, letting users express intent naturally with no onboarding. However, text struggles to convey structured data that is common in business, and without real-time updates, static text results lose relevance the moment they’re generated.

Structured data is easier to digest when users can filter, sort, and visualize it—that is why graphical user interfaces (GUIs) excel at presenting structured data and guiding users through complex workflows. But GUIs are expensive to build and rigid, forcing generic, one-size-fits-all solutions that struggle to provide the fluid, tailored experiences users now demand.

Text is flexible but limited; GUIs are robust but rigid. Generative UI is the unmet need between them and the new frontier for business software.

From static dashboards to dynamic workspaces

Imagine a procurement manager investigating a supply chain disruption. Instead of navigating five different applications and manually cross-referencing data, she asks: “Show me the suppliers at risk in Southeast Asia and model alternative sourcing scenarios.”

This request sets agents to work behind the scenes. They gather and analyze live data, simulate outcomes, and calculate the projected impact of every alternative. Execution agents are also pre-positioned and ready to act on command.

The user doesn’t have to deal with any of this complexity. For them, a dynamic interface materializes in seconds—not a generic dashboard, but a purpose-built mission control center. Interactive maps highlight affected regions and supply chain graphs update in real time. As the user tweaks parameters, risk scores adjust instantly. Embedded controls stand ready to trigger purchase orders or notify suppliers, enabling the user to decide and execute. Collaboration is simplified; colleagues can join a living workspace: no briefing decks, no context-setting calls.

This is the future: a business suite where a user’s intent defines their interface and their decisions drive action. To get there, we are combining Joule and Joule Agents with our vision for generative UI. This is not just about on-demand dashboards; it’s about steering a business with interfaces that adapt to each user’s role, context, and tasks. This is “vibe coding” for enterprise operations: shifting focus from syntax to intent.

We are entering an era where AI constructs UIs on the fly, allowing users to engage with them immediately. Generative UI marks the transition from static software suites to “batch size 1” applications that act like ephemeral control centers tailored to a specific problem.

Challenges and SAP’s answers

Delivering an intent-driven business suite at enterprise scale requires addressing complex realities. We are building generative UI because we understand its promise and its perils—and we have unique assets to bridge that gap.

Accuracy

Large language models (LLMs) can produce plausible but incorrect outputs, or “hallucinate.” A consumer chatbot that hallucinates a movie plot is tolerable; a procurement system that misrepresents supplier terms has real consequences. Our generative UI approach addresses this by visualizing data directly from systems of record with transparent lineage. Grounding the UI in real-time, trusted data is our first defense against inaccuracy.

Trust

If every interface is generated on the fly, how do users know it is reliable? Trust is built on consistency and predictability. Our generative UI is built on the familiar and proven architectural grammar of SAP Fiori for lists, dashboards, and workflows. The content is bespoke and the structure is consistent and familiar, so users can always judge and adjust with confidence.

Complexity

Enterprise systems are sophisticated and unique. They are built over decades, encoding massive domain knowledge and business logic. Generative UI builds on Joule’s existing integration and orchestration capabilities, which already connect to systems across a landscape and coordinate agents to execute complex workflows. Generative UI leverages this foundation, letting users interact with deeply integrated processes through simple interfaces while Joule handles the orchestration underneath.

Why this matters now

The expectations set by consumer AI are real, and the gap between what employees experience at home and what they use at work is widening.

The future of enterprise software isn’t chatbots bolted onto legacy screens. It’s bespoke mission control—interfaces that materialize around a user’s intent, grounded in live data, executed by agents, and governed by the user.

With that, we’re reimagining how work gets done.

Jonathan von Rueden is chief AI officer of SAP SE.