Giotto.ai, a Swiss AI research lab founded in 2017 by researchers at EPFL (the Swiss federal technology university in Lausanne), announced a collaboration with SAP on June 8 to explore integrating advanced reasoning capabilities into SAP Joule Agents. The partnership will focus on pilot projects deploying Giotto.ai’s compact reasoning models in enterprise workflows that require structured decision-making and deep integration with business data.
What the Partnership Covers
The collaboration targets a specific gap in enterprise agent architectures: reasoning under constraints. While frontier models from OpenAI and Anthropic optimize for general intelligence, Giotto.ai has built what it calls a “portable reasoning layer,” compact models designed to run inside private enterprise environments rather than relying on centralized cloud infrastructure.
SAP will pilot Giotto.ai’s technology in scenarios where Joule Agents need to make structured decisions within complex business processes. According to Yaad Oren, Head of SAP Research and Innovation, the work “directly supports our vision of trusted and governed agentic AI solutions that can operate reliably within complex business processes.”
Giotto.ai CEO Aldo Podesta framed the move as a validation step: “We are entering a phase where enterprise AI must go beyond basic use cases into more ambitious ones that require advanced reasoning and (very) large context understanding,” he told Finanzwire.
Why SAP’s Agent Stack Matters
The timing aligns with SAP’s broader push into agentic AI. At SAP Sapphire in late May, the company announced Joule Studio, a fully managed platform for building and managing enterprise AI agents. Joule Studio integrates with SAP’s Knowledge Graph, Business Data Cloud, and Domain Models, giving agents native access to live business data. Sony reported that Joule Studio “generated an end-to-end solution in 10 to 15 minutes, replacing three to four days of manual development,” according to SAP’s announcement.
Giotto.ai’s reasoning models slot into this stack as a specialized inference layer. Rather than replacing frontier models entirely, the pitch is that enterprise agents need purpose-built reasoning for tasks like procurement decisions, compliance checks, and supply chain optimization, where accuracy and auditability matter more than creative generation.
The Portable Reasoning Bet
Giotto.ai is positioning itself as an alternative to the U.S. and Chinese frontier model labs, emphasizing Swiss neutrality and European data sovereignty. The company is also working on a capital raise “to consolidate its independent and sovereign model and serve companies and institutions across Switzerland and Europe,” according to the press release.
The approach reflects a growing split in enterprise AI: frontier labs compete on raw capability, while a second tier of companies builds specialized, deployable reasoning for specific domains. For agent builders, the question is whether compact, on-premises reasoning models can deliver reliability that general-purpose APIs cannot, particularly in regulated industries where sending business data to external endpoints creates compliance risk.
SAP’s 400,000+ enterprise customer base gives Giotto.ai a distribution channel that most AI startups cannot access independently. Whether the pilots produce measurable results will determine if “portable reasoning” becomes an infrastructure category or remains a niche.