Reviewer comparing printed billing pages with a notebook at a wooden desk

Flagship

Compute cost review

A three-week reading of one AI application’s invoices, usage windows, and idle capacity, delivered as a findings memo you can take to finance.

Ask about this review

Who it is for

Product owners, engineering leads, and finance partners in Malaysia whose teams already run an AI feature — chat, retrieval, transcription, image generation, or a similar call pattern — and need a written account of where last quarter’s compute money went.

What you leave with

A findings memo covering line items, unused capacity, and a simple allocation table. You leave with numbers you can defend in a budget meeting, not a slide deck of general advice.

Scope

One named application, one billing account, and the most recent ninety days of invoices and usage extracts. Cross-account sprawl belongs in a later review.

Duration
Typically three weeks from a complete invoice pack to the memo. Incomplete logs stretch the calendar; we will say so at intake rather than guess.
Where
Kuala Lumpur rooms by appointment, video for teams outside the city, or on-site at your KL office when the invoice pack is large enough to warrant it.
Fee basis
From RM 12,400 for a single-application review. Extra billing accounts, extra regions, or a second application are quoted after intake.
Who leads
Farah Idris leads the review. Lim Wei Han joins when usage logs need mapping to product surfaces. Sessions are held from our rooms on Jalan Thambi Dollah or by video.

Included

  • Kickoff call to name the application, the billing account, and the people who hold invoices
  • Collection and reading of three months of compute invoices and a sample of usage logs
  • Two working sessions (on video or at our Kuala Lumpur rooms) to walk line items together
  • A written findings memo with idle-capacity notes and a first-pass allocation table
  • One follow-up call within ten working days of the memo

Not included

  • Rewriting application code or changing model routing
  • Negotiating credits or contracts with your cloud or model vendor
  • Ongoing monitoring, alerting, or a standing dashboard
  • Reviews of more than one billing account in the same engagement

How the sitting runs

  1. Intake

    You send the last three months of compute invoices, a one-page sketch of the application, and the name of the person who can answer billing questions.

  2. Reading

    We mark GPU-hour lines, token charges, reserved versus on-demand capacity, and anything that looks idle after a traffic spike.

  3. Working sessions

    Two sessions, ninety minutes each. We sit with the marked invoices. You correct what we have misread.

  4. Memo

    You receive a findings memo and a one-page allocation table. We walk it once on a follow-up call.

Preparation

Bring three months of invoices (PDF or CSV), a named billing owner, and a rough picture of which product surfaces call the model. We do not need production credentials at intake. If a later step needs a read-only usage export, we will write the exact grant on paper first.

Limits we keep

We do not hold production keys. We do not keep your invoices after the follow-up call unless you ask us to retain them for a later briefing. Reviews pause if the billing account name on the invoice does not match the application you asked us to read.

Next step

Write to us with the application name and the months you want read. We reply within two working days with whether the pack is enough to start.

Request this review