What engineering leaders and their teams report about GitHub Copilot, with every finding tied to its evidence.
Read the findingsPractitioner sentiment, leaders and teams
+15 Mixed from 20 sources that take a side
How this figure is madeFrom 20 authors in 19 discussions. Small sample. Use a task pilot to measure accepted work and review time.
Jan 2025 to Oct 2026.
72 sources from practitioners, by month and sentiment.
Leader and organization accounts split between gains and problems. They agree that a gain needs training and champions on top of the licenses; Indeed retired Copilot when use stalled despite both. The concern they raise most is proving impact to the business, followed by the review Copilot output needs.
Positive, mixed and negative sources
The score counts each source once, as the split does. See How the numbers work.
See the evidenceBelow: 2 praise themes, 5 concerns, 1 close to agreement and 1 disagreement.
Counted for Engineering leaders and teams, from 41 sources. Change group
Every capability practitioners describe positively, ranked by how many sources say so. Open one to read who said it.
Counted for Engineering leaders and teams.
Positive, 5
Positive
David Griffiths, Chief Technology Officer, Citi
American Banker: Citi is rolling out agentic AI to its 40,000 developersNews report quoting the CTO
Citi's CTO estimates regular Copilot users get 5% to 15% more productivity and calls Copilot a good assistive tool, while Devin takes simple, repeatable tasks.
""We still very much believe in GitHub Copilot," Citi's Chief Technology Officer David Griffiths told American Banker. "Copilot is a good assistive tool for developers. It can help them complete tasks faster, because it effectively auto-completes parts of those tasks for them."…"
Positive
Ali Dasdan, with Gal Bakal, Guy Levin, Michael Kaufman, and Yaniv Katz, ZoomInfo Engineering
Experience with GitHub Copilot for Developer Productivity at ZoominfoEngineering blog post, ZoomInfo Engineering Blog
ZoomInfo rolled Copilot out to over 400 developers in four phases and measured a 33% suggestion acceptance rate, 20% of lines accepted, 72% satisfaction, and about 20% time saved.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep
Mixed
Yaroslav (Yarik), tech lead for the coding-agent rollout, Superhuman
How Superhuman Scaled AI Coding Agents Across Their Entire Eng OrgConference talk video
Superhuman's engineers started with Copilot's autocomplete, which they found useful, then moved to Cursor and Claude Code for agent work.
"…a year and a half ago, we first experienced Copilot. It was very cool auto complete."
Mixed
James Camilleri, Lead Developer at Scott Logic
Rapid web app development with Devin - A Developer’s PerspectiveEngineering blog post, Scott Logic blog
Switching between Copilot and Devin on one app, found Copilot needed tight prompts, limited context and line-by-line review like a junior's work. Advises small pull requests so AI code does not swamp reviewers.
"When working with Copilot I found I needed to rein it in to get good output. Given too much freedom it would overproduce inappropriate code and go down paths I wanted to avoid."
Practitioner account, name and role on their organization's own page; direct use; counts toward Delivered outcomes
Positive
Birlla Packiasamy, Thoughtworks
Our experiment with GitHub Copilot: A practical guide for development teamsEngineering blog post, Thoughtworks
Over a 10-week experiment a four-developer team went from 20 to 23 stories per sprint, after a slow start and a drift away from test-driven development they had to correct.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep
Mixed or negative, 1
Positive
Government Digital Service, Department for Science, Innovation and Technology
AI coding assistant trial: UK public sector findings reportGovernment trial report, GOV.UK (DSIT and GDS)
A three-month cross-government trial with 1,100 redeemed Copilot Business licenses reported 56 minutes saved per day, a 15.8% line acceptance rate, and 39% of users committing suggested code.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep
Counted for Engineering leaders and teams.
Positive, 2
Positive
K3UL, Hacker News commenter
Comment on: Kimi K2.7 Code is generally available in GitHub CopilotHacker News comment
Their company compared costs after the June change and kept Copilot: a seat includes matching usage at provider prices across several model vendors, unlike Claude Enterprise.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Positive
Ali Dasdan, with Gal Bakal, Guy Levin, Michael Kaufman, and Yaniv Katz, ZoomInfo Engineering
Experience with GitHub Copilot for Developer Productivity at ZoominfoEngineering blog post, ZoomInfo Engineering Blog
ZoomInfo rolled Copilot out to over 400 developers in four phases and measured a 33% suggestion acceptance rate, 20% of lines accepted, 72% satisfaction, and about 20% time saved.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep
Mixed or negative, 4
Negative
qxxx, Hacker News commenter
Comment on: How many products does Microsoft have named 'Copilot'?Hacker News comment
Their boss told developers paying for personal Copilot to use the company's Copilot, which turned out to be the Microsoft 365 product, a different tool from GitHub Copilot.
Practitioner account, pseudonymous; second-hand; not counted, takes no position
Mixed
parkersweb, Hacker News commenter who manages a GitHub organization
Comment on: Updates to GitHub Copilot interaction data usage policyHacker News comment
Managing an organization that licenses Copilot for some developers, they received the data policy email with no mention that Business seats were excluded.
Practitioner account, pseudonymous; ran the rollout; not counted, takes no position
No opinion stated
hades200082, GitHub Community member writing for their company
The #1 reason my company won't use Github Copilot right nowGitHub Community discussion, Copilot Conversations
Their company will not roll Copilot out because assigning a company seat cancels a developer's personal Copilot subscription on the same account, which ties side projects to company tooling and IP terms.
"When our company assigns a Copilot seat to a developer, GitHub automatically cancels their individual Copilot subscription on that same account."
Practitioner account, pseudonymous; opinion; not counted until substantiated
Mixed
kookamamie, Hacker News commenter
Comment on: GitHub Copilot Coding AgentHacker News comment
Their organization on GitHub Business could not work out how to enable the coding agent, and they want tighter control than write access over who can assign it work.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Business and Enterprise disable the cloud agent and third-party MCP servers until an organization enables them. Organizations can block agent access to selected repositories or all repositories. Adding Copilot cloud agent to your organization
Open them in the recordEvery concern practitioners raise, most raised first.
Counted for Engineering leaders and teams.
Raised as a concern, 4
Mixed
James Camilleri, Lead Developer at Scott Logic
Rapid web app development with Devin - A Developer’s PerspectiveEngineering blog post, Scott Logic blog
Switching between Copilot and Devin on one app, found Copilot needed tight prompts, limited context and line-by-line review like a junior's work. Advises small pull requests so AI code does not swamp reviewers.
"When working with Copilot I found I needed to rein it in to get good output. Given too much freedom it would overproduce inappropriate code and go down paths I wanted to avoid."
Practitioner account, name and role on their organization's own page; direct use; counts toward Delivered outcomes
No opinion stated
Hao Li, Haoxiang Zhang, and Ahmed E. Hassan, Queen's University
The Rise of AI Teammates in Software Engineering (SE) 3.0: How Autonomous Coding Agents Are Reshaping Software EngineeringResearch paper, arXiv (Queen's University)
Studying agentic pull requests in the AIDev dataset (456,535 PRs), the authors report OpenAI Codex with the highest acceptance rate at 64%, then Devin at 49% and GitHub Copilot at 35%. Accepted Copilot PRs took 17.2 hours to review, over 4 times the 3.9 hours for accepted human PRs.
"In contrast, GitHub Copilot shows the opposite pattern: slow to accept, fast to reject. Accepted GitHub Copilot PRs take 17.2 hours to review, which is over 4 times longer than accepted Human-PRs (3.9 hours)…"
Practitioner account; structured test; counts toward Delivered outcomes
Mixed
Ostap Elyashevskyy, Test Automation Competence Manager at ELEKS
AI Code Review in Practice: Expert Evaluating GitHub CopilotCompany engineering blog post, ELEKS
ELEKS found Copilot code review good at naming, unused imports and common best practices, but it missed a null check and stays at surface-level feedback that cannot replace human review.
"GitHub Copilot (as of now) tends to focus more on surface-level suggestions, like syntax, naming, formatting, and common best practices, rather than deep code review feedback. It cannot substitute human code review."
Practitioner account, name and role on their organization's own page; direct use; counts toward Delivered outcomes
Positive
Ali Dasdan, with Gal Bakal, Guy Levin, Michael Kaufman, and Yaniv Katz, ZoomInfo Engineering
Experience with GitHub Copilot for Developer Productivity at ZoominfoEngineering blog post, ZoomInfo Engineering Blog
ZoomInfo rolled Copilot out to over 400 developers in four phases and measured a 33% suggestion acceptance rate, 20% of lines accepted, 72% satisfaction, and about 20% time saved.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep
Discussed without concern, 2
Positive
Angie Jones, maintainer of goose, Block's open source AI agent
How I Taught GitHub Copilot Code Review to Think Like a MaintainerPersonal blog post, angiejones.tech
goose maintainers found Copilot code review noisy, with about one comment in five a real catch, until repository instructions on confidence, priorities and CI coverage cut the noise.
Practitioner account, self-described; ran the rollout; counts toward Delivered outcomes, Adoption and upkeep
Positive
Birlla Packiasamy, Thoughtworks
Our experiment with GitHub Copilot: A practical guide for development teamsEngineering blog post, Thoughtworks
Over a 10-week experiment a four-developer team went from 20 to 23 stories per sprint, after a slow start and a drift away from test-driven development they had to correct.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.A Copilot pull request requires another reviewer's approval because the requester's approval does not count. Actions workflows also wait for human approval. Review output from Copilot
Open them in the recordCounted for Engineering leaders and teams.
Raised as a concern, 4
Mixed
Viktoria Stray, University of Oslo and SINTEF, with Elias Goldmann Brandtzæg, Viggo Tellefsen Wivestad, Astri Barbala, and Nils Brede Moe
Developer Productivity With and Without GitHub Copilot: A Longitudinal Mixed-Methods Case StudyResearch paper, arXiv (University of Oslo and SINTEF)
Analyzing 26,317 commits from 25 Copilot users and 14 non-users at NAV IT over two years, the authors found no statistically significant change in commit activity after adoption, though minor increases appeared. Developers mainly used Copilot for boilerplate, documentation and repetitive tasks.
"We did not find any statistically significant changes in commit-based activity for Copilot users after they adopted the tool, although minor increases were observed."
Practitioner account; structured test; counts toward Delivered outcomes
Negative
insane_dreamer, Hacker News commenter
Comment on: GitHub Copilot is moving to usage-based billingHacker News comment
The billing announcement showed their small team had been paying for Copilot through its GitHub bill while nobody used it, so they canceled it.
Practitioner account, pseudonymous; ran the rollout; not counted, repeats a stronger account
Negative
koyote, Hacker News commenter
Comment on: GitHub Copilot is moving to usage-based billingHacker News comment
At a Microsoft shop, data governance makes switching providers hard, and they expect the price change to make Copilot ROI negative and cut AI use in the company.
Practitioner account, pseudonymous; opinion; not counted, takes no position
Negative
isqueiros, Hacker News commenter
Comment on: Changes to GitHub Copilot individual plansHacker News comment
Their company holds company-wide Copilot, Windsurf and ChatGPT Enterprise accounts, yet nearly everyone uses Claude Code through Bedrock instead.
Practitioner account, pseudonymous; second-hand; not counted, takes no position
Discussed without concern, 8
Positive
Ty Smith, Principal Engineer at Uber
Uber’s journey of measuring AI impact on developer productivityConference session page with transcript, DX Annual via Engineering Enablement by DX
Smith reports that when Copilot and Cursor launched, Uber lacked usable telemetry, so it ran qualitative surveys before layering in causal methods. Co-presenter Abhishek Tibrewal reports a difference-in-differences study found a 19% incremental lift in PRs per engineer for Copilot power users, about 8% of engineers.
"The core question was, if there was no Copilot, what would have the same developer shipped? And that’s your counterfactual, and the delta between that counterfactual and what the developers actually shipped is the incremental impact that we were looking at. We got 19% incremental impact for this power user cohort."
Practitioner account, self-described; ran the rollout; counts toward Delivered outcomes
Mixed
Kelly Anne Pipe, Head of Developer Experience at Vanguard
Augmented, accelerated, autonomized: How Vanguard is embedding AI across the product lifecycleConference session page with transcript, DX Annual via Engineering Enablement by DX
Vanguard rolled Copilot out to its engineers and found pockets of power users, but faster coding alone did not speed delivery because the rest of each product team still worked at the old pace.
"We had rolled out Copilot for engineers. There were some pockets of brilliance. We had some power users doing really incredible things."
Practitioner account, self-described; ran the rollout; counts toward Delivered outcomes
Mixed
Jason Valentino, Head of Software Engineering Strategy at BNY
Mapping the new SDLC at BNY: Codifying AI into every step of the delivery lifecycleConference session page with transcript, DX Annual via Engineering Enablement by DX
Valentino reports that BNY rolled out Copilot in 2024 and brought the first strong use cases back to its community of practice. Once coding assistants were everywhere, BNY was "not exactly" making more money, so it looked across the whole delivery lifecycle; it now uses Windsurf, Copilot and Claude Code.
"It’s 2024, we’re rolling out Copilot. As the first couple use cases show up where someone’s like, “Wow, this is really good at this.” Then come back to the community of practice, get those on there."
Practitioner account, self-described; ran the rollout; not counted, takes no position
Positive
Government Digital Service, Department for Science, Innovation and Technology
AI coding assistant trial: UK public sector findings reportGovernment trial report, GOV.UK (DSIT and GDS)
A three-month cross-government trial with 1,100 redeemed Copilot Business licenses reported 56 minutes saved per day, a 15.8% line acceptance rate, and 39% of users committing suggested code.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep
Positive
Birlla Packiasamy, Thoughtworks
Our experiment with GitHub Copilot: A practical guide for development teamsEngineering blog post, Thoughtworks
Over a 10-week experiment a four-developer team went from 20 to 23 stories per sprint, after a slow start and a drift away from test-driven development they had to correct.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep
Positive
Ali Dasdan, with Gal Bakal, Guy Levin, Michael Kaufman, and Yaniv Katz, ZoomInfo Engineering
Experience with GitHub Copilot for Developer Productivity at ZoominfoEngineering blog post, ZoomInfo Engineering Blog
ZoomInfo rolled Copilot out to over 400 developers in four phases and measured a 33% suggestion acceptance rate, 20% of lines accepted, 72% satisfaction, and about 20% time saved.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep
Mixed
sandos, Hacker News commenter
Comment on: GitHub Copilot is moving to usage-based billingHacker News comment
Has Copilot at work and personally. On their own and nearby teams, about one in three people use it much, so pooled credits depend on a few heavy users.
Practitioner account, pseudonymous; second-hand; not counted, takes no position
Positive
David Betts, leader of the developer platform team at Twilio
Leveraging sentiment data, driving org-wide action, and executive engagementPodcast episode page with transcript, Engineering Enablement by DX
Betts cites Twilio engineers self-reporting that GitHub Copilot saves them two to four hours a week, which Betts frames as four full-time engineers of time saved across the organization, and says high-frequency Copilot users correlate with high sentiment on a variety of other drivers.
"…people are self-reporting that they’re saving two to four hours by using GitHub Copilot per week, and that equates to four full-time engineers of time saved across our organization."
not counted until substantiated
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.The usage dashboard measures IDE activity. It omits Copilot Chat on GitHub.com, GitHub Mobile, code review, and Copilot CLI. The impact dashboard groups users by adoption and connects those groups to pull request throughput. Viewing the Copilot impact dashboard
Open them in the recordCounted for Engineering leaders and teams.
Raised as a concern, 3
Mixed
Cem Dilmegani, with Şevval Alper, AIMultiple
AI Code Review Tools BenchmarkIndependent benchmark, AIMultiple
Replaying 309 pull requests, AIMultiple ranked Copilot code review third behind CodeRabbit and Greptile; its suggestions were accurate and good at catching typos, with shorter analysis than the two leaders.
"GitHub Copilot is very successful at finding typos in code and makes spot-on suggestions; its analysis is shorter than those of CodeRabbit and Greptile."
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep
Negative
kasey_junk, Hacker News commenter
Comment on: Kimi K2.7 Code is generally available in GitHub CopilotHacker News comment
When vendor management at work assumes Claude through Copilot equals Claude Code, their side-by-side runs show Copilot underperforming three or four times out of five.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
No opinion stated
Hao Li, Haoxiang Zhang, and Ahmed E. Hassan, Queen's University
The Rise of AI Teammates in Software Engineering (SE) 3.0: How Autonomous Coding Agents Are Reshaping Software EngineeringResearch paper, arXiv (Queen's University)
Studying agentic pull requests in the AIDev dataset (456,535 PRs), the authors report OpenAI Codex with the highest acceptance rate at 64%, then Devin at 49% and GitHub Copilot at 35%. Accepted Copilot PRs took 17.2 hours to review, over 4 times the 3.9 hours for accepted human PRs.
"In contrast, GitHub Copilot shows the opposite pattern: slow to accept, fast to reject. Accepted GitHub Copilot PRs take 17.2 hours to review, which is over 4 times longer than accepted Human-PRs (3.9 hours)…"
Practitioner account; structured test; counts toward Delivered outcomes
Discussed without concern, 2
Mixed
James Camilleri, Lead Developer at Scott Logic
Rapid web app development with Devin - A Developer’s PerspectiveEngineering blog post, Scott Logic blog
Switching between Copilot and Devin on one app, found Copilot needed tight prompts, limited context and line-by-line review like a junior's work. Advises small pull requests so AI code does not swamp reviewers.
"When working with Copilot I found I needed to rein it in to get good output. Given too much freedom it would overproduce inappropriate code and go down paths I wanted to avoid."
Practitioner account, name and role on their organization's own page; direct use; counts toward Delivered outcomes
Positive
Ali Dasdan, with Gal Bakal, Guy Levin, Michael Kaufman, and Yaniv Katz, ZoomInfo Engineering
Experience with GitHub Copilot for Developer Productivity at ZoominfoEngineering blog post, ZoomInfo Engineering Blog
ZoomInfo rolled Copilot out to over 400 developers in four phases and measured a 33% suggestion acceptance rate, 20% of lines accepted, 72% satisfaction, and about 20% time saved.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep
Counted for Engineering leaders and teams.
Raised as a concern, 3
Mixed
Kon5ole, Hacker News commenter
Comment on: Kimi K2.7 Code is generally available in GitHub CopilotHacker News comment
A fan of Copilot CLI and its model switching whose whole department moved to Claude Code after the June 2026 billing change burned the monthly budget every two to three days.
Practitioner account, pseudonymous; direct use; counts toward Operational control, Cost predictability and limits
Negative
insane_dreamer, Hacker News commenter
Comment on: GitHub Copilot is moving to usage-based billingHacker News comment
The billing announcement showed their small team had been paying for Copilot through its GitHub bill while nobody used it, so they canceled it.
Practitioner account, pseudonymous; ran the rollout; not counted, repeats a stronger account
Negative
koyote, Hacker News commenter
Comment on: GitHub Copilot is moving to usage-based billingHacker News comment
At a Microsoft shop, data governance makes switching providers hard, and they expect the price change to make Copilot ROI negative and cut AI use in the company.
Practitioner account, pseudonymous; opinion; not counted, takes no position
Discussed without concern, 3
Positive
K3UL, Hacker News commenter
Comment on: Kimi K2.7 Code is generally available in GitHub CopilotHacker News comment
Their company compared costs after the June change and kept Copilot: a seat includes matching usage at provider prices across several model vendors, unlike Claude Enterprise.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Mixed
sandos, Hacker News commenter
Comment on: GitHub Copilot is moving to usage-based billingHacker News comment
Has Copilot at work and personally. On their own and nearby teams, about one in three people use it much, so pooled credits depend on a few heavy users.
Practitioner account, pseudonymous; second-hand; not counted, takes no position
Mixed
on_the_train, Hacker News commenter
Comment on: GitHub Copilot is moving to usage-based billingHacker News comment
At a large semiconductor company Copilot is the only approved LLM tool and approving another would take years, so the company will pay the new price.
Practitioner account, pseudonymous; opinion; not counted, takes no position
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Spending limits send alerts unless Stop usage when budget limit is reached is enabled. User-level budgets always stop usage. Unexpected charges alone do not show that controls are absent. Getting started with budget controls
Open them in the recordCounted for Engineering leaders and teams.
Raised as a concern, 2
Negative
kevcampb, Hacker News commenter
Comment on: Updates to GitHub Copilot interaction data usage policyHacker News comment
The 2026 interaction data policy change pushed their company into checking vendor by vendor that its code is not used to train models.
not counted, repeats a stronger account
No opinion stated
hades200082, GitHub Community member writing for their company
The #1 reason my company won't use Github Copilot right nowGitHub Community discussion, Copilot Conversations
Their company will not roll Copilot out because assigning a company seat cancels a developer's personal Copilot subscription on the same account, which ties side projects to company tooling and IP terms.
"When our company assigns a Copilot seat to a developer, GitHub automatically cancels their individual Copilot subscription on that same account."
Practitioner account, pseudonymous; opinion; not counted until substantiated
Discussed without concern, 3
Negative
koyote, Hacker News commenter
Comment on: GitHub Copilot is moving to usage-based billingHacker News comment
At a Microsoft shop, data governance makes switching providers hard, and they expect the price change to make Copilot ROI negative and cut AI use in the company.
Practitioner account, pseudonymous; opinion; not counted, takes no position
Mixed
parkersweb, Hacker News commenter who manages a GitHub organization
Comment on: Updates to GitHub Copilot interaction data usage policyHacker News comment
Managing an organization that licenses Copilot for some developers, they received the data policy email with no mention that Business seats were excluded.
Practitioner account, pseudonymous; ran the rollout; not counted, takes no position
Mixed
kookamamie, Hacker News commenter
Comment on: GitHub Copilot Coding AgentHacker News comment
Their organization on GitHub Business could not work out how to enable the coding agent, and they want tighter control than write access over who can assign it work.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
GILD's fact checks, which don't count as sources
Checked Oct 28, 2026.GitHub excludes Business and Enterprise data from model training. Those plans carry IP indemnity, which since April 3, 2026 no longer requires the public code filter, per Microsoft's Customer Copyright Commitment page. Free, Pro, and Pro+ interactions permit training unless users opt out. GitHub Copilot plans & pricing FAQ
Open them in the recordStatements this group's sources agree on, then the ones close to agreement.
Close to agreement: each needs 9 supporting sources
Copilot adoption requires training and internal champions, plus time to learn. Organizations that only distribute licenses report uneven use.
8 supporting sourcesQuestions the sources split on, with the count on each side.
Evidence remains mixed
Main tool, 3 sources
These engineers keep Copilot as their main tool for editor integration or model choice and price. Others use it because their employer approves it.
Moved to another tool, 7 sources
These engineers moved their main work to Claude Code, Cursor, Codex, or another agent. Some retain Copilot for completions. Others drop it.
The three groups, then every practitioner by the context they work in, and how emphasis changes with scale.
Three groups drawn from the same sources. Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. Signals open on Engineering leaders and teams.
41 sources. These accounts weigh organization-wide rollouts, enablement, measurement, data and IP terms, spend under AI credits, and what changes when Copilot is the approved default across an engineering organization.
31 sources. These accounts weigh completions, agent mode, the cloud agent and Copilot CLI, editor integration, and how Copilot's pricing compares with the alternatives.
72 sources. This reading is the broadest, and it mixes solo and organization settings that can reach very different conclusions.
72 sources from practitioners, by the context they work in. Filled rows count for this group.
Emphasizes Agent quality and pricing when switching tools
Emphasizes Sprint throughput and paying for unused seats
Emphasizes Admin and data policies, plus review noise
Emphasizes IP and terms, coaching, and what to measure
Emphasizes Acceptance and time savings with the approved tool
What changes with scale. Individual builders compare Copilot's agents and prices with Claude Code, Cursor, and Codex. Leaders and organizations focus on procurement and training, plus data terms and rollout measurement.
Where this group's sources were published.
Counted for Engineering leaders and teams, from 41 sources. Change group
63 sources for this group, 22 of them documentation
GILD keeps older evidence that still applies. Accounts tied to premium requests, earlier plans, or retired models carry Needs a recheck.
See sources by originWho each source is, how it knows, whether it still applies and what it counts toward.
Newest first, on its own page. Every count above opens its sources there.
Read all 94 sourcesAbout this read
A running review of what engineering teams report about GitHub Copilot, from individual builders to company-wide rollouts. Each finding links to its sources, and each conflict shows both sides.
Counts cover only the sources GILD reviewed.
Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. GILD files each source in a group from how the account describes its author. A role or employer marked self-described is what the source states about themselves. Documentation checks facts for every group and never counts toward sentiment.
A source's date and whether it still applies are separate: a recent post can describe an old version, and an old one can still hold.
Flagged for a recheck, of the sources that take a side: leaders and teams 0 of 20, individual builders 10 of 29, all practitioners 10 of 49.
Each row opens its sources for the group you are reading.
| Finding | Engineering leaders and teams | Individual builders | All practitioners | Sources |
|---|---|---|---|---|
| Current readCurrent sentiment | 7 positive, 9 mixed, 4 negative | 12 positive, 11 mixed, 6 negative | 19 positive, 20 mixed, 10 negative | Sources |
| PraiseInline Completions | 5 of 6 positive | 4 of 4 positive (under the minimum) | 9 of 10 positive | Sources |
| PraiseAgent Execution | 1 of 3 positive (under the minimum) | 6 of 12 positive | 7 of 15 positive | Sources |
| PraiseEditor Integration | 0 of 0 positive (under the minimum) | 6 of 6 positive | 6 of 6 positive | Sources |
| PraiseGitHub Integration | 2 of 4 positive (under the minimum) | 7 of 7 positive | 9 of 11 positive | Sources |
| PraiseEnterprise Administration | 2 of 6 positive | 0 of 0 positive (under the minimum) | 2 of 6 positive | Sources |
| PraiseCode Quality | 0 of 3 positive (under the minimum) | 5 of 7 positive | 5 of 10 positive | Sources |
| PraiseValue for Money | 1 of 3 positive (under the minimum) | 4 of 6 positive | 5 of 9 positive | Sources |
| ConcernUsage / Cost | 3 of 6 raise it | 4 of 11 raise it | 7 of 17 raise it | Sources |
| ConcernRequires Strong Review | 4 of 6 raise it | 1 of 4 raise it (under the minimum) | 5 of 10 raise it | Sources |
| ConcernNeeds Close Steering | 1 of 1 raise it (under the minimum) | 2 of 6 raise it | 3 of 7 raise it | Sources |
| ConcernFeature Lag | 0 of 0 raise it (under the minimum) | 1 of 3 raise it (under the minimum) | 1 of 3 raise it (under the minimum) | Sources |
| ConcernQuality Against Alternatives | 3 of 5 raise it | 6 of 7 raise it | 9 of 12 raise it | Sources |
| ConcernData, IP & Security | 2 of 5 raise it | 0 of 0 raise it (under the minimum) | 2 of 5 raise it | Sources |
| ConcernMeasuring Impact | 4 of 12 raise it | 0 of 0 raise it (under the minimum) | 4 of 12 raise it | Sources |
| ConsensusLicenses alone do not produce the gain | 8 supporting (below threshold) | 0 supporting (under the minimum) | 8 supporting (below threshold) | Sources |
| ConsensusLargest gains on routine code and tests | 3 supporting (under the minimum) | 1 supporting (under the minimum) | 4 supporting (under the minimum) | Sources |
| ConsensusWritten instructions improve results | 2 supporting (under the minimum) | 2 supporting (under the minimum) | 4 supporting (under the minimum) | Sources |
| ConsensusChosen as the approved default | 4 supporting (under the minimum) | 2 supporting (under the minimum) | 6 supporting (below threshold) | Sources |
| DisagreementIs Copilot the main AI coding tool, or the one engineers keep beside another? | Main tool 3, Moved to another tool 7 | Main tool 7, Moved to another tool 8 | Main tool 10, Moved to another tool 15 | Sources |
| DisagreementDoes Copilot's usage-based billing work for the people paying for it? | Workable 1, Unpredictable 2 (under the minimum) | Workable 6, Unpredictable 7 | Workable 7, Unpredictable 9 | Sources |
| DisagreementOnce review and rework are counted, does Copilot reduce engineering time? | Net time saved 5, Absorbed by review 0 (under the minimum) | Net time saved 1, Absorbed by review 0 (under the minimum) | Net time saved 6, Absorbed by review 0 (under the minimum) | Sources |
| Current readEvidence mix | 63 sources | 53 sources | 94 sources | Sources |
GILD's checks of claims against official documentation, kept apart from the source counts.
