Introduction
Many enterprises have treated artificial intelligence as a series of isolated experiments. That approach is reaching its limits as AI agents move into core business workflows. Leaders are now being asked to integrate AI into the operating model itself, rather than bolting it onto existing processes. In a discussion produced by MIT Technology Review Insights in partnership with Infosys, two SAP executives outlined what it takes to move from pilot projects to scaled, trustworthy AI across the organization.
What Happened
Thorsten Leiduck, SVP and global head of SAP Business AI Platform Ecosystem, and Satinder Pal Singh, associate VP for North America at SAP Business AI Platform and SAP AI Strategy at Infosys, described a fundamental shift in how companies should approach AI. Instead of launching disconnected proofs of concept, organizations should treat AI as an enterprise platform built on a connected, semantically rich data layer.
Leiduck emphasized that AI performance depends entirely on the quality of the underlying data. While 80 percent accuracy may be acceptable for drafting emails or summarizing meetings, it is insufficient for mission-critical processes such as financial close or supply chain management.
Singh introduced the idea of AI agents as "digital coworkers." He recommended evaluating potential use cases through three lenses: business impact, technical feasibility, and scalability. He warned that a pilot unable to scale remains little more than an experiment.
The conversation also highlighted research showing that most companies have not moved beyond early-stage AI work. A BCG report cited in the discussion found that 74 percent of organizations are stuck in pilot mode with no scaling plan. Separate research from Infosys and HFS indicated that only 14 percent of Global 2000 enterprises have successfully scaled AI, while an Infosys Knowledge Institute index found that just 8 percent of companies qualify as AI native.
Why This Matters
As AI takes on more responsibility, governance becomes a competitive issue rather than a compliance formality. Clear policies, access controls, monitoring, and oversight are needed to keep agents aligned with business goals. Singh noted that the absence of governance does not accelerate adoption; it creates a risk backlog that can force a disruptive stop-and-audit later.
The executives also pointed to a growing divide between early movers and laggards. Companies classified as AI native generate four times the revenue contribution from AI compared with others, according to the Infosys Knowledge Institute. That advantage compounds over time through agent behavior data, calibrated knowledge graphs, and workforce familiarity.
Key Takeaways
- AI should be treated as an enterprise platform, not a collection of isolated projects.
- A strong, semantically rich data foundation is a prerequisite for reliable AI at scale.
- Use cases should be selected based on impact, feasibility, and scalability.
- AI agents work best when viewed as digital coworkers embedded in business processes.
- Governance, including access controls and observability, must be built in from the start.
- Organizations that delay scaling risk falling behind competitors that are already building compounding AI advantages.
Conclusion
Moving AI from the lab to the enterprise operating model requires more than new technology. It demands a unified data strategy, clear governance, and a mindset shift that treats AI as a shared organizational capability. For companies ready to move past the pilot phase, the path forward involves starting with high-value use cases, proving ROI, and scaling what works—all while ensuring agents remain trustworthy, observable, and aligned with business outcomes.

Discussion
Join the conversation
Thoughtful reactions, questions, and follow-up ideas help shape the next story.