A real screenshot of the GPT interface, showing the assistant as it was shared with the FinOps community rather than relying on a generic thumbnail.
In August 2025, Joey Wilkes and I built and shared a custom GPT for the FinOps community after hearing recurring adoption and onboarding friction around FOCUS. The goal was simple: use the public material already available at focus.finops.org and the specification PDF to create a free, low-friction way for practitioners to ask questions and learn faster.
What We Heard
In a FOCUS contributor member meeting, the adoption problem was clear: people wanted to contribute, teach, and practice against the standard, but the barrier to entry was still too high. The official material was strong, but not always easy for newcomers to interrogate conversationally. We treated that as a practical product problem, not just a documentation problem.
What We Built
- A custom GPT in ChatGPT trained against publicly available FOCUS content and the spec PDF.
- A free community-facing proof of concept shared directly with the FinOps community.
- A lightweight way for practitioners to ask questions about columns, requirements, contribution guidelines, maintainers, and version changes.
- A testbed for deciding whether a more deliberate, self-hosted or better-trained approach was worth pursuing.
Community-first
Built in response to adoption friction discussed in the member meeting.
Free to use
Shared as a public GPT for anyone with a ChatGPT account, with rate limits on free plans.
Spec-grounded
Based on public FOCUS material and the PDF specification rather than invented summaries.
Feedback-driven
Released openly so community members could test it, break it, and report confusion.
Launch Positioning
We shared it with a direct ask to the community: try it, post your best and worst prompts, and tell us where it failed. That mattered. The point was not to declare victory; it was to validate whether conversational access could help onboarding and standards adoption.
Product Visual
Community Feedback
The early responses were useful because they were specific. The feedback showed both the promise and the weaknesses of a first-pass GPT built on public spec material.
- Cory Syvenky (Teck): said it would have been helpful while preparing a FinOps X presentation, then tested version-difference and dataset-validity prompts that exposed confusion between FOCUS 1.1 and 1.2 context.
- My response: this was a proof of concept to determine whether a more specific, more controlled training and testing path was worth pursuing.
- Arvind Shastry: responded that it was an effective and very cool way to learn about FOCUS.
Prompts We Tested
- Tell me about the columns, their description, and the normative requirements of the 1.1 specification.
- Who were the maintainers of the 1.1 specification?
- What are the contribution guidelines?
- What are the normative requirements?
- What is the main difference between the 1.0 specification and the 1.1 specification?
What It Proved
- There is real demand for conversational, low-friction onboarding around FOCUS.
- Version-awareness and source-scoping are the hard parts in standards-oriented GPT work.
- A GPT can accelerate learning, but only if its grounding and testing strategy are rigorous.
- Community feedback is the fastest way to find where a semantic tool is genuinely useful and where it is still brittle.
Why This Matters Professionally
This project shows the operating pattern I care about: listen to the friction, build a practical artifact quickly, release it to real users, and use their feedback to learn where a better system needs to be more precise. It also sits directly alongside my FOCUS knowledge-base work and the semantic-layer thread behind my FinOps X 2026 talk.
Related Work
This GPT was one part of a larger effort to make FOCUS easier to learn and use. The adjacent project is my FOCUS Knowledge Base, which structures the standard as a semantic reference system rather than a flat document set.