Harness launched two new products on May 28 targeting what it calls the defining challenge of enterprise AI in 2026: proving that AI spend produces measurable outcomes.

The products AI DLC Insights and Cloud & AI Cost Management are now in beta. 

The launch is grounded in Harness's own research: a survey of 700 engineering practitioners across five countries found that 94% of engineering leaders say the metrics that matter most are missing from their current frameworks.

The problem breaks into two surfaces:

• Developer side: Engineers write nearly every line of new code with AI. But token spend has never been connected to outcomes. Leaders can't see what fraction of AI-generated code ships or how much spend is wasted on abandoned code.
• Infrastructure side: Agents in production consume tokens continuously. In most organizations, spending is visible only at the invoice level. A $28,000 monthly spend on a customer support agent is a completely different number depending on how many tickets it resolved. Most organizations cannot tell the difference today.

Harness CEO Jyoti Bansal called one symptom "tokenmaxxing", an engineer burns 500,000 tokens generating code that gets rejected in review, and by the leaderboard, they outperform the engineer who shipped a clean 50-line patch.

What the two products do

AI DLC Insights runs an on-machine agent in the developer's environment, captures every AI-generated line of code with its token cost, and maps it through the delivery chain to the PR, ticket, and deployment it produced.

Cloud & AI Cost Management provides a unified view of spend across every AI provider with attribution traced to the agent, model, team, and business unit. Anomaly detection flags spend spikes before the invoice arrives.

For organizations managing both sanctioned AI infrastructure and shadow AI adoption across the broader SaaS estate, engineering-layer visibility alone isn't enough. 

CloudEagle.ai provides the complementary layer, tracking AI tool usage and token consumption across the full application footprint, including unsanctioned tools that platform-specific solutions don't cover.

‍

Your AI Stack Has Blind Spots

Uncover every one.

Get The Guide