Production case study · Nick Cerutti
LLM cost reduction program
$200K/yr aggregate reduction in cloud and LLM spend.
Portfolio of client platforms across the advisory engagements. Cost optimization program · FinOps.
Context and scope
The $200K/year figure is aggregate cloud and LLM spend reduction across advisory engagements.
The same integration mistakes repeated across client codebases: context re-injected without caching, token-heavy integrations, and infrastructure patterns that multiplied spend.
What shipped
- Refactored Anthropic SDK, n8n, and Copilot Studio integrations around token optimization, prompt caching, and low-infrastructure patterns.
Outcomes
- $200K/yr aggregate reduction in cloud and LLM spend across the portfolio.
Sources and scope
author reported · Reviewed 2026-09-22
Client outcomes. Most client identities and raw operational records are not public; Softograph is named, and the remaining clients stay anonymous. $200K/year is aggregate savings across engagements. $40K/day was projected exposure; 72 hours describes that incident.
What to check in your system
The common pattern was integration overhead: repeated context and token-heavy execution paths. The work focused on those paths and their infrastructure requirements. Estimating a new project's savings still requires its traffic mix, model prices and quality constraints.
- How much repeated context is sent on each request?
- Which tasks require the current model and which have alternatives?
- How will you check quality and latency alongside a lower bill?
Stack: Anthropic SDK · n8n · Copilot Studio · AWS Bedrock
Working on a similar problem?
Three weeks: find where the money goes, fix the largest lines, and leave the controls that stop it coming back.
LLM Cost & Routing Sprint · $15K-20K