# Lanai Releases Token Tuner To Reduce Token Spend > Source: > Published: 2026-05-27 18:20:21.492731+00:00 # Lanai Releases Token Tuner To Reduce Token Spend The New Stack reports that "Tokenmaxxing is real, expensive & it's spreading," and highlights new tooling to curb rising AI API bills. According to The New Stack, Lanai's Token Tuner maps token spend to specific workflows and surfaces opportunities to substitute lower-cost models for premium ones. Lanai's public product pages show a dashboard that detects model usage and workflow adoption across an organization, listing detected tools such as Claude, ChatGPT, Gemini, Copilot, and Cursor and metrics like **847** people and **68%** workforce adoption (Lanai product pages). A VP OPERATIONS quoted on Lanai's site said, "We knew AI was saving time. We did not know where the leverage actually was until Lanai showed us." Editorial analysis: this class of observability and model-routing tooling addresses a growing FinOps pain point for teams running multi-model, multi-workflow deployments. ### What happened The New Stack runs with the headline "Tokenmaxxing is real, expensive & it's spreading," and reports on a wave of tools aimed at reigning in skyrocketing AI API costs (The New Stack). According to The New Stack, **Lanai's Token Tuner** maps token spend to individual workflows and identifies where lower-cost models can replace premium ones (The New Stack). Lanai's product pages present a dashboard that enumerates detected AI tools, workflow adoption rates and usage counts; the site lists **847** people detected and **68%** workforce AI adoption in example dashboards (Lanai product pages). A VP OPERATIONS quoted on Lanai's site said, "We knew AI was saving time. We did not know where the leverage actually was until Lanai showed us" (Lanai product pages). ### Technical details Per Lanai's product pages, the Token Tuner surfaces token consumption mapped to workflows and detected assistants, plus approval status for deployed agents (Lanai product pages). The dashboard examples show detected models and tools including Claude, ChatGPT, Gemini, Copilot, and Cursor, and call out counts of approved versus unapproved AI uses (Lanai product pages). The public material frames the capability as visibility at the workflow level - attributing spend to use cases rather than solely to projects or teams - and highlighting substitution opportunities where lower-cost models can satisfy the same workflow requirements (The New Stack; Lanai product pages). ### Editorial analysis - technical context Companies operating multi-model stacks and agentic workflows increasingly confront per-call and per-token cost leakage driven by long contexts, model choice, and unmanaged assistants. Industry-pattern observations: teams facing those pressures typically adopt three levers, observability, model routing (policy-based selection), and prompt or cache optimization, to reduce spend without wholesale feature rollback. ### Context and significance Editorial analysis: For ML engineers and platform teams, tools that translate token usage into workflow-level signals change where cost controls are applied. Rather than tuning individual prompts or negotiating price alone, platform observability that highlights high-volume workflows enables targeted routing to cheaper models, staged caching, and workload-specific SLAs. ### What to watch Editorial analysis: Watch for integrations between token-level observability and model-rerouting/traffic-splitting systems, native billing connectors to verify realized savings, and vendor support for multi-model policy enforcement. Also track whether similar features appear in MLOps platforms and cloud provider tools, which would broaden adoption and standardize metrics. ## Scoring Rationale This is a practical product-level development with clear relevance to ML platform and FinOps practitioners. It is not a frontier-model release, but it addresses a rising operational pain that affects real deployment costs. Practice interview problems based on real data 1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with. [Try 250 free problems](/problems)