---
title: "The new economics of AI cost management"
description: "Enterprise AI cost management is shifting from maximizing raw token volume to maximizing return on investment, requiring teams to measure, attribute, and optimize spend across models, agents, and applications. Granular cost attribution, enabled through request-level tags that map spend to specific teams or projects, is essential because organizations cannot manage what they do not measure. To avoid context bloat, systems should have agents pull relevant context on demand via tools or MCP servers rather than pushing all information upfront, preventing unnecessary token waste. Dynamic model routing assesses task complexity in real time to direct the 90% of tasks suited for cheaper models away from frontier models. End-to-end tracing collects the full exhaust of agent activity including prompts and tool calls, enabling teams to identify abuse, waste, and leaks to inform cost management decisions."
---

# The new economics of AI cost management

[Databricks](https://yomu.fyi/company/databricks) · Databricks Staff · Oct 22, 2026

**Type:** Explainer

## Summary

Enterprise AI cost management is shifting from maximizing raw token volume to maximizing return on investment, requiring teams to measure, attribute, and optimize spend across models, agents, and applications. Granular cost attribution, enabled through request-level tags that map spend to specific teams or projects, is essential because organizations cannot manage what they do not measure. To avoid context bloat, systems should have agents pull relevant context on demand via tools or MCP servers rather than pushing all information upfront, preventing unnecessary token waste. Dynamic model routing assesses task complexity in real time to direct the 90% of tasks suited for cheaper models away from frontier models. End-to-end tracing collects the full exhaust of agent activity including prompts and tool calls, enabling teams to identify abuse, waste, and leaks to inform cost management decisions.

## Context

Interview with Kevin Stumpf, Senior Staff Engineer at Databricks and Technical Lead for Unity Gateway, discussing the shift from tokenmaxxing to valuemaxxing and practical AI cost management through attribution, routing, guardrails, and tracing.

## Approach / What changed

Implement AI cost management through granular cost attribution using request-level tags, dynamic model and harness routing based on task complexity, on-demand context retrieval to avoid bloat, spend guardrails with hard caps and budgets, and end-to-end tracing of agent workflows.

## Takeaways

- Granular cost attribution using request-level tags allows organizations to trace AI spend to specific users, teams, or projects and answer where spend comes from and what it produced.
- Context bloat can be avoided by letting agents pull relevant context on demand through tools or MCP servers rather than pushing all information upfront.
- Dynamic routing should assess task complexity to route 5-10% of tasks to frontier models and the rest to cheaper alternatives, and harness routing further optimizes cost by controlling agent behavior and context management.

## FAQ

### What is context bloat and how do you fix it?

Context bloat means pushing excessive information to an agent, wasting tokens and distracting it. The fix is not generating the bloat: let the agent pull relevant context on demand through tools, MCP servers, or skills rather than pushing everything upfront.

### How does dynamic model routing work for cost optimization?

Smart routing assesses task complexity in real time and decides whether to use a frontier model or cheaper alternative. Only 5-10% of tasks genuinely need frontier models; the rest can use lower-cost models while maintaining quality.

### Why is end-to-end tracing important for AI cost management?

Tracing collects the full exhaust of agent activities – prompts, tool calls, data – enabling identification of waste, abuse, security leaks, and patterns. More traces lead to more insight for managing costs.

**Tags:** [AI](https://yomu.fyi/topic/ai), [Cost Optimization](https://yomu.fyi/topic/cost-optimization), [Enterprise AI](https://yomu.fyi/topic/enterprise-ai), [Model Routing](https://yomu.fyi/topic/model-routing), [Tracing](https://yomu.fyi/topic/tracing)

- Source: [Databricks](https://www.databricks.com/blog/ai-cost-management)
- Source URL: https://www.databricks.com/blog/ai-cost-management
- Ingested by Yomu: 2026-10-22T21:00:11.497Z

[Read original post](https://www.databricks.com/blog/ai-cost-management)
