Panel, Conference Presentation
WEKA, Intel, Graphite, InstaDeep & AWS: Scaling Without Crashing. AI Framework for Enterprise Growth
- The AI adoption trajectory is expected to evolve from 2023 experiments and last year's production moves to scaling "agent swarms" with dozens of parallel instances, shifting strategy toward the inference market where approximately 80% of AI spending is projected to go.
- Businesses are predicted to transition from simple API calls to deploying agent swarms that deliver software and projects to market faster and more robustly, requiring the aggregation of disparate context data from tools like Slack, Jira, and Linear.
- Mid-sized companies with revenues between €250 million and €1 billion are allocating significant investments to position AI at the center of transformation, establishing dedicated ex-co level roles for governance and clear data transparency frameworks to ensure trust and regulatory compliance.
- Economic outcomes are projected to diverge significantly by 2025, with AI companies expected to become profitable entities and economic expansion widening between organizations adopting positive AI cycles and those that do not, driven by compute as the primary differentiator as models reach near-genius intelligence levels.
- Infrastructure and operational models must adapt to engineers writing three to ten times more code, shifting the developer's role from writing lines to prompting and reviewing, with pull requests and code review interfaces becoming the central point of development despite the potential decline of "vibe coding" branding.
- Specific performance and cost targets include a 75% discount on cached input pricing, SLM deployment on edge NPUs in PCs, and Intel's Xeon 6 delivering 50% more inference performance for one-third of the code volume.
- Sector-specific predictions include biotech and health tech advancements outpacing previous capabilities within five years, with agent groups in diagnostics expected to significantly outperform doctors, while personalized medicine generation relies on new insights into previously ignored DNA.
- Organizations face risks regarding the shift from data-centric to compute-centric models, security vulnerabilities from centralizing previously obscure context data, and the potential for short-term investment pressures to hinder the transition toward longer-term strategies and open, interoperable solutions.
- Future AI governance is expected to incorporate distributed designs focusing on ethics and humanity, with the SPARC framework recommended for orchestrating agent swarms and open source cited as essential to prevent hardware vendor locking.