Enterprise AI Shifts From Experimentation to Functional Scale
Enterprise AI is shifting from experimental pilots to scalable functional transformation, according to David Robinson, president of SAP North America. Companies are increasingly deploying AI agents to automate complex business processes and reduce modernization risks, moving away from traditional feature-based product roadmaps toward interoperable, autonomous systems.
Robinson told InformationWeek that after speaking with customers at the SAP Sapphire event in May, he observed a distinct change in corporate strategy. Organizations are no longer simply “placing bets” on artificial intelligence; they’re now focusing on how those investments translate into scalable innovation and tangible business value.
How AI agents are changing business operations
AI agents have matured enough to handle specific business processes, which Robinson says could fundamentally reshape corporate strategy. Unlike basic chatbots, these agents analyze information, provide recommendations, and make proactive decisions with limited human intervention based on established inputs.
This shift is altering how companies plan their technology growth. Robinson suggests that traditional, feature-by-feature product roadmaps are becoming less critical. In their place, enterprises are prioritizing “interoperability and the ability to have trusted, skilled agents” to shorten delivery times and streamline production cycles.
What is the SAP Autonomous Enterprise?
SAP is currently integrating AI more deeply with physical operations through its Autonomous Enterprise initiative. This framework combines AI agents with a platform designed for data orchestration and workflows.
To illustrate the practical application, Robinson cited a retail scenario where the system manages the end-to-end process of designing and producing personalized, monogrammed hats for consumers. This moves AI out of the realm of theoretical productivity and into the actual execution of physical goods and services.
Why model curation matters more than model power
The current trend in the enterprise market isn’t a search for a single, “best” AI model. Instead, companies are realizing the necessity of curating a variety of different models for specific use cases across their ecosystem.
“It’s not how good the models are, but how well you curate all of those models and are able to adapt,” Robinson said. This approach allows companies to match the specific strengths of a model—such as reasoning, speed, or creativity—to the specific business requirement it’s meant to solve.
How AI reduces IT modernization costs
AI is already altering IT operating models by reducing the cost, risk, and time associated with modernization work. Robinson noted that many CTOs and CIOs are currently resetting their AI agendas as more features become available.
A key prerequisite for this future is the consolidation of legacy systems. Many companies are evaluating how to maintain older, customized enterprise resource planning (ERP) environments to ensure they can actually take advantage of autonomous resources. Without this cleanup, the “leap” into autonomous operations remains impossible.
Agent-Led vs. Feature-Led Development
The transition from feature-led to agent-led development represents a significant shift in how software is consumed in the enterprise.
| Feature-Led Approach | Agent-Led Approach |
|---|---|
| Focuses on adding specific tools or buttons to a UI. | Focuses on achieving an outcome via autonomous reasoning. |
| Requires manual user navigation through menus. | Uses orchestration to trigger workflows across systems. |
| Roadmaps defined by a list of new capabilities. | Roadmaps defined by interoperability and agent skill sets. |
Common Questions About Enterprise AI Transition
Are companies fully replacing human workers with AI agents?
No. Robinson described the current phase as “moderate strides rather than full leaps.” Enterprises are still determining the exact balance of how much AI will change the way IT and business operations function.
Why can’t companies just use one powerful LLM for everything?
Different tasks require different optimizations. Curation allows a business to use a lightweight model for simple data entry and a high-reasoning model for complex strategic analysis, optimizing both cost and performance.
What is the biggest barrier to AI adoption in the enterprise?
Legacy infrastructure. According to Robinson, consolidating older systems is a prerequisite for utilizing autonomous resources effectively.