Enterprises Will Optimize AI Models with Custom Flywheels

@AravSrinivas· June 27, 2026 View original

Summary

The post predicts that every enterprise will develop its own AI model development and evaluation system, focusing on "token value per watt optimization." This is driven by the need to leverage unique tacit knowledge about their domain and customers.

The author posits that the future of enterprise AI involves each company establishing a bespoke "model-harness-sandbox-eval flywheel." This integrated system will be designed to optimize AI models for "token value per watt," emphasizing efficiency and performance. The core rationale behind this trend is the unique, tacit knowledge that individual enterprises possess regarding their specific domains, customer behaviors, and internal workflows, which cannot be fully captured by general-purpose models.

Why it matters

This vision suggests a shift towards highly customized and efficient AI deployments, requiring professionals to build internal capabilities for model development, evaluation, and optimization tailored to their specific business context.

How to implement this in your domain

  1. 1Begin developing an internal framework for AI model experimentation and evaluation.
  2. 2Invest in talent and tools for fine-tuning and optimizing models with proprietary data.
  3. 3Establish metrics for "token value per watt" or similar efficiency measures for AI operations.
  4. 4Document and integrate unique domain knowledge into AI training and validation processes.
  5. 5Explore hybrid AI strategies combining off-the-shelf models with custom components.

Who benefits

All industriesTechFinanceHealthcareManufacturing

Key takeaways

  • Enterprises will increasingly build custom AI development and evaluation pipelines.
  • Optimization will focus on efficiency metrics like "token value per watt."
  • Unique domain and customer knowledge will be a key differentiator for enterprise AI.
  • This approach enables companies to leverage their proprietary insights effectively.

Original post by @AravSrinivas

"Every enterprise will have its own model-harness-sandbox-eval flywheel with token value per watt optimization. This is the future. Simple reason: tacit knowledge about the domain and customers and their workflows that the company uniquely understands and has built trust around."

View on X

Originally posted by @AravSrinivas on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses