GitHub Copilot logo

[ AI tools · Development ]

GitHub Copilot, Used Well Inside Your Policies

We work with Copilot in your GitHub organization, from completions and agent mode to the coding agent that turns issues into pull requests, and we set it up so your engineers get the same results.

Fix double-counted discounts on refunds #1192

Draft
  1. Issue #1186 assigned to CopilotTotals double-count discounts on refunds
  2. Commit: add failing test for refund totalscopilot · 2 files
  3. Commit: apply discount once in refund pathcopilot · 1 file
  4. Commit: update receipt snapshotcopilot · 1 file
  • Build
  • Unit tests
  • CodeQL
  • Lint

Copilot review: change is limited to the refund path; new test covers the bug. Review requested from @priya (senior engineer).

Example pull request from Copilot coding agent.

From the Editor to the Pull Request

Copilot runs on models from OpenAI, Anthropic and Google, chosen per task, and lives where your code and reviews already are.

[ Copilot coding agent ]

What Happens After You Assign an Issue

The coding agent works asynchronously, like a teammate who picks up a ticket. These five steps are where we configure and check it.

[ Plans and controls ]

What Each Copilot Plan Lets You Control

A simplified view for planning. GitHub’s plan page has the current limits and prices; we check them with you before rollout.

  • Individual
    • Completions, chat and agent mode
    • Coding agent and code review: Pro and up
    • Choice of models
  • Business
    • Completions, chat and agent mode
    • Coding agent and code review
    • Choice of models
    • Organization policies and model controls
    • Content exclusion for sensitive paths
    • IP indemnity
    • Usage metrics and audit logs
  • Enterprise
    • Completions, chat and agent mode
    • Coding agent and code review
    • Choice of models
    • Organization policies and model controls
    • Content exclusion for sensitive paths
    • IP indemnity
    • Usage metrics and audit logs
    • Enterprise-wide policy and managed users
[ What to assign ]

Issues the Coding Agent Handles Well

Clear scope, a test that proves it and an owner who will review it. These are the kinds of issues we assign to Copilot on client projects.

  • #1201

    Raise test coverage in a module

    Add tests for the edge cases in refunds.ts until coverage passes 85%.

    Assignee: Copilot · Reviewer: code owner

  • #1202

    Fix accessibility findings

    Resolve the contrast and label issues from the latest audit report.

    Assignee: Copilot · Reviewer: code owner

  • #1203

    Small, well-described bugs

    Fix the date shown in the wrong time zone on invoices, with a test.

    Assignee: Copilot · Reviewer: code owner

  • #1204

    Dependency and API updates

    Move from the deprecated payments endpoint to the new one.

    Assignee: Copilot · Reviewer: code owner

  • #1205

    Documentation

    Write usage examples for every public function in the SDK.

    Assignee: Copilot · Reviewer: code owner

  • #1206

    Technical debt with clear rules

    Replace callback-style code in the jobs folder with async functions.

    Assignee: Copilot · Reviewer: code owner

Laptop and wide monitor showing lines of code on a desk
Rollout log01 / 04.github/copilot-instructions.md committedOrg policy: models, exclusions, MCP allowlistPilot: review time and lead time trackedSession: agent mode on ticket #482
[ Rollout ]

Copilot Across an Engineering Organization

Four steps we take before a whole organization switches on Copilot.

  1. 01

    Instructions in every repository

    Copilot instruction files and AGENTS.md with commands, conventions and off-limits paths.

  2. 02

    Policies set once

    Allowed models, content exclusion, coding agent repositories and MCP servers configured at the organization.

  3. 03

    A pilot that measures

    Two or three teams with goals for review time and lead time, tracked with Copilot usage metrics.

  4. 04

    Training on real tickets

    Sessions where engineers use agent mode and the coding agent on their own backlog, not demos.

Draw a Card

Three ways to start. Pick one to bring it to the front. Packages from $1,500; custom projects from $8,000.

Copilot, Asked and Answered

  • GitHub states that code from Copilot Business and Enterprise is not used to train its models. Organization admins also control content exclusion, allowed models and which repositories the coding agent can work in.

[ Contact ]

Let's build your next product together

Book a free strategy call and leave with a clear plan and estimate. No commitment.

  1. 01Pick a time that suits you
  2. 0230 minutes on scope, stack, timeline and budget
  3. 03Fixed-price proposal, NDA on request
Prefer email? contact@sajaltech.com