Learning to Fit the Room: A Skill I Built to Practice Concision

I once received feedback that I shared too much detail when explaining something. Even when passionate about a topic, it doesn’t automatically mean I communicate it in the way someone else needs. Sometimes, a conversation requires just the essentials, while other times, more is necessary. Engineers may focus on trade-offs, while product teams consider end-users, and management seeks an overview of progress and decisions needed.

My native language is Spanish, and most professional communication happens in English, adding another layer to this challenge. The core issue isn’t just finding the right words, but choosing which details matter in the moment. Concision emerged as a way to practice being more concise. The goal was to consistently apply the advice “Be more concise,” which is easier said than done.

An AI assistant can shorten a message, but shorter isn’t always clearer. A quick Teams update requires a concise conclusion and next step, while a technical document must provide enough detail for someone else to understand the work. A spoken explanation should sound like something I’d actually say, rather than a lifted paragraph from a report.

Using the framework from Harsha Perera’s article, The Art of Being Concise, I developed Concision as a portable SKILL.md for AI agents. It helps edit a message, explain a topic, offer feedback, or work through a draft. The underlying principle remains the same: first understand the purpose and the person, then decide how much to say.

The value lies in consistency, ensuring this principle carries through meetings, teams updates, documentation, or any work in progress. To demonstrate, I created a performance pilot using a support application with a slow ticket search. Engineering is testing a cache to improve performance. In one controlled staging comparison, p95 search latency dropped from 820 to 510 milliseconds, while the error rate remained at 0.2%.

I then crafted three illustrative drafts for different audiences using Concision’s principles. To engineering, in Teams, a quick update would be: "Quick update on the search cache: in our staging test, p95 latency dropped from 820 ms to 510 ms with the cache enabled. Errors stayed at 0.2%. Security still needs to check tenant isolation and access to cached results, and QA has permission changes and stale results to cover.

Nothing’s approved for production. Engineering will send the design to Security on Tuesday." To product, in a conversation, the focus would be on the need to make ticket search faster for support staff. The staging test showed p95 response time improved from 820 to 510 milliseconds, with errors staying at 0.2%. However, real user impact remains unknown, and security and QA reviews are pending before deciding on a limited rollout test.

To management, an email would summarize the promising staging test results, noting that security and QA reviews are still open, and more representative performance testing is needed before recommending a rollout. Production approval has not been requested yet. Engineering will send the design to Security on October 13, and they’ll share a recommendation after the remaining checks.

These are fictional examples, but the key takeaway is the ability to tailor messages for different audiences while maintaining the core information. Over time, I aim to make these decisions more naturally, using AI to support the practice while remaining responsible for the content I convey. I’m sharing Concision with the hope that it will help others navigate the complexities of communicating across teams, providing a framework to structure ideas and determine what details are essential for each conversation.

The skill is available on GitHub under the MIT license, with installation instructions and examples for various AI agents. You can install it using npx skills add franciscoh017/concision.