Conference Presentation, Panel, Fireside Chat
Alix, SAP, Daytona, Feedly, Prosus & Handelsblatt: The Future Agentic AI & Autonomous Workflows
RAISE SummitAlix, Edwin Kodabachan, Paul Van der Boor, Alexander Maisor, Emmanuel Casimates, Ivan Bourazin
- Foundational models are expected to evolve from simple request-response interactions to supporting multi-step planning, reasoning, and end-to-end workflows by connecting to various applications.
- Agents are predicted to interact with the world sequentially, moving from APIs to unstructured environments like browsers, then entire computer systems, and eventually real-world robots.
- A shift in business models is anticipated where individuals and startups can build agents on platforms like SAP, potentially creating an economy where agents autonomously consume services and data from providers.
- Software developers' output is projected to double within the next two years, while automation of mundane tasks is expected to free humans for creative work, with a specific expectation that agents will handle 85% to 90% of complex tasks.
- Human oversight will remain necessary for high-complexity tasks such as those in wealth management, whereas simple, repeated tasks with well-defined parameters can be completed autonomously without human intervention.
- Agentic quality controls are expected to enable agents to progress from "intern" to "senior" levels based on data labeling and evaluation, with new roles emerging where humans manage digital teammates and oversee quality control.
- Security and governance frameworks are becoming critical as customers request controls to prevent agents from accessing unauthorized ports, starting or stopping processes, or making unauthorized purchases, with limits needed on spending and actions.
- A significant expansion of bureaucracy and "governance monster" entities is expected to manage on-prem, hybrid, cloud, and sovereign environments, requiring strict logging of agent actions to determine authentication and authorization.
- The definition of "agentic" remains uncertain regarding security, privacy, and autonomy, leading to varying implementation challenges as regulations and controls are developed to enable future agent communication.
- Interface paradigms will likely shift from traditional applications to dynamic, AI-native environments where voice interaction and deeper reasoning improve experiences, though the specific form factor remains undefined.
- Infrastructure and API standards will need to be established to allow agents to navigate existing platforms efficiently, moving away from web navigation toward direct service requests (e.g., Domino's Pizza API) to handle exponential traffic growth.
- Large organizations leveraging their data and processes will gain a speed advantage over competitors, as the pace of innovation is projected to exceed previous historical rates, with the intelligence-per-dollar trend continuing to decline.
- While individual capabilities could theoretically enable the creation of billion-dollar companies with specific skill sets, this is not considered the norm; instead, value extraction from agents is expected to continue for the next decade even without "step function" model improvements.
- The transition from human-centric to human-agent collaboration involves a premium on human interface roles, with companies needing to monitor agent behavior through board reviews and ensure right guardrails are in place for exponential progress.