The Token Economy: How AI architecture shapes cost and utility

Token costs are exploding—and your AI architecture, not just your models, will determine whether those costs turn into real business value. Learn how Glean and Swiggy design for high token yield: more useful, work-ready output per token. What you’ll learn: • How “token yield” works and why enterprise AI cost and utility are being shaped by architecture decisions, not models alone • Why deep enterprise context (indexing vs. federation, knowledge graphs, MCP tools) dramatically improves both quality and token efficiency • How intelligent routing, model choice, and new harness patterns (sub-agents, coding harness, compaction) reduce token waste at scale • Real-world lessons from Swiggy’s AI-native concierge and how they use Glean as a context layer to power efficient, governed AI systems Timestamps: 00:00 Why enterprise AI token costs are skyrocketing 02:30 Token yield explained: outcomes per token 06:10 Glean’s enterprise AI architecture: context, routing, harness, learning 14:45 Swiggy’s approach to context, metrics, and AI-native products 24:30 Practical strategies for optimizing AI architectures and evaluations Subscribe for more: https://www.youtube.com/@gleanwork Follow us: • LinkedIn: https://www.linkedin.com/company/gleanwork/ • Twitter/X: https://x.com/glean • Instagram: https://www.instagram.com/gleanwork/