The 376% Forrester figure is real. It measures GitHub Enterprise Cloud as a whole, not the return on your Copilot rollout. A CFO-ready ROI case needs your costs, adoption, delivery, and quality data.
GitHub commissioned Forrester Consulting to examine the potential financial impact of GitHub Enterprise Cloud. Forrester interviewed five decision-makers at four organizations and built a composite organization with 5,000 developers. The study calculated $85.9 million in three-year benefits against $18.1 million in costs, producing a 376% ROI and payback in less than six months.
Those findings are useful, but the scope matters. The quantified benefits include developer productivity from GitHub Enterprise Cloud and Copilot, security and compliance efficiencies, faster onboarding, retirement of legacy tools, and revenue from faster time to market. The number supports the broader GitHub platform, not a return directly from Copilot seats.
Your CFO is asking a narrower question: what incremental value does Copilot produce across teams at your current adoption and usage levels after the full cost of the deployment?
GitHub's analytics provide much more than simple seat reports. Its documented metrics include active users, agent adoption, acceptance rates, lines changed with AI, AI credits consumed, pull request creation and merge counts, and median time to merge.
The remaining gap is outcome and financial attribution. GitHub's metrics don't calculate net ROI or show whether Copilot-touched code increased rework or production incidents. Our AI business case framework explains why operating activity still has to be connected to quality, cost, and business value.
The August 19, 2026 analysis of Microsoft’s AI coding agent field study adds real-world output evidence to the enterprise investment conversation. Researchers used direct telemetry from tens of thousands of engineers during Microsoft’s early-2026 rollout of Claude Code and GitHub Copilot CLI, rather than relying only on developer surveys or isolated lab tasks.
As reported in that analysis, the study estimated that adopters merged 24% more pull requests per engineer per day than they would have without the tools. The 95% confidence interval ranged from 14.5% to 33.7%. Researchers found no statistically meaningful decline between February and March–April, supporting persistence across the roughly four-month observation window, not a guarantee of lasting gains beyond it.
The study covers one organization’s rollout of two agentic command-line tools. Its headline result is not a standalone return on GitHub Copilot, not evidence about every Copilot workflow, and not a 24% financial ROI gain. Do not use it as an automatic productivity assumption for your enterprise.
A merged pull request is an output measure, not a feature shipped, a customer problem solved, or durable code in production. The study does not establish whether the additional output improved quality, reduced incidents or rework, or created more business value. The paper identifies quality as an open question.
The Forrester and Microsoft findings therefore answer different questions. Forrester models the financial impact of a broader GitHub Enterprise Cloud investment. Microsoft estimates a change in merged PR output during an agent-tool rollout. Neither replaces a Copilot-specific assessment of your incremental costs and accepted business outcomes.
The study’s cost warning, summarized in the linked analysis, is that token consumption can reach millions of dollars annually at organizational scale. That is a risk to measure, not a typical Copilot cost estimate. The paper’s extreme spending illustration comes from a separate Fortune report about a Meta employee, not a Microsoft engineer or a representative participant in the study.
For your rollout, evaluate delivery, quality, and cost over the same period. More output can justify higher consumption when it reaches production and remains useful. The same output lift can produce weak economics if retries, review, defects, and rework absorb the apparent gain. Attribute costs to tools, teams, engineers, agents, and use cases where the data supports it; retain unknown allocations rather than manufacturing precision.
Start with delivery and quality data from before Copilot adoption: lead time, deployment frequency, change failure rate, rework, and code turnover. Then track how those measures change as adoption deepens.
A simple before-and-after comparison can still be misleading if the work mix, team composition, or release cycle changes. Compare similar teams, repositories, and work types where possible, and document assumptions used to reconstruct a missing baseline.
An enterprise-wide active-user rate can hide teams that barely use Copilot and power users generating most of the activity and cost. Measure adoption depth by team and role, then connect it to the work those teams deliver.
GitHub prices Copilot Business at $19 per user per month and Copilot Enterprise at $39. Both include monthly GitHub AI Credits, with usage-based charges above those allowances. Copilot Enterprise requires GitHub Enterprise Cloud, but you should only include that license in the ROI calculation if the Copilot purchase creates an additional platform cost for your organization.
Larridin's AI Adoption dashboard fills the utilization visibility gap by showing adoption depth by team and role. Token Spend & Insights attributes seat and usage costs across tools, teams, and workflows. Together, they connect adoption with the cost side of the ROI model.
Acceptance rate shows that an engineer took a suggestion. It doesn't show whether the code merged, reached production, or held up after 30 or 90 days.
Connect Copilot usage to pull request throughput, lead time, change failure rate, incidents, rework, and code turnover. Larridin's AI Dev Productivity platform combines GitHub and Jira data with AI contribution and quality signals, helping leaders see where Copilot improves delivery and where apparent gains shift work downstream.
For agent-assisted work, keep three aligned views: output such as complexity-adjusted throughput and cycle time; quality such as code durability, review effort, rework, failures, and incidents; and cost including consumption and the human effort needed to verify and remediate results. Preserve merged PR counts as a supporting signal, not the productivity score. Account for task splitting and work complexity before interpreting higher counts as additional capacity.
The cost side of the ROI calculation should include seats, AI credit overages, implementation, enablement, governance, review, and remediation. The benefit side can include faster delivery, avoided outside spend, reduced overtime, fewer defects, or engineering capacity redirected to measurable priorities.
Don't convert every hour saved into cash automatically. Forrester's GitHub Enterprise Cloud model applies a 50% productivity recapture assumption to reflect that saved time becomes financial value only when the organization uses it productively. A defensible Copilot model should make that conversion explicit.
The 376% figure is accurate for Forrester's composite GitHub Enterprise Cloud organization over three years. It isn't a standalone measurement of GitHub Copilot. The model includes benefits and costs from the broader platform, including Actions, Advanced Security, tool consolidation, and Copilot.
GitHub's usage and impact metrics cover adoption, engagement, acceptance rate, lines of code, agent activity, pull request lifecycle, and AI credit consumption. They can show how Copilot is used and how usage relates to pull request output. A complete ROI case still needs production quality, rework, full cost, and financial value data.
It provides several inputs, but it doesn't produce a net ROI calculation. Leaders still need to connect Copilot usage and cost to delivery, quality, and business outcomes, then apply an agreed financial model.
Copilot Business is $19 per user per month and Copilot Enterprise is $39, with monthly AI credit allowances and usage-based overages. Include the incremental cost of GitHub Enterprise Cloud for Copilot Enterprise only when that platform purchase is part of the Copilot decision. Also include enablement, governance, integration, review, and remediation costs.
No. The study estimates merged-PR output during Microsoft’s rollout of Claude Code and Copilot CLI. It does not establish a financial return, business-value increase, or a standalone Copilot effect. Use it to inform measurement design, then test your own baseline, quality, full cost, and accepted results.
The source analysis reports no statistically meaningful decline between the later periods compared within the roughly four-month study window. That supports persistence during observation, not a prediction that gains will continue indefinitely.
No. The source analysis says observation ended April 29, 2026, and an internal announcement shortly afterward directed most affected engineers to move to Copilot CLI. The study does not establish that cost or its productivity findings caused the decision. Do not treat the sequence as proof of comparative ROI.
Larridin connects Copilot usage and spend to delivery, quality, and cost outcomes at the team and business-unit level, closing the gap between native operating metrics and a board-ready ROI case.
Book a discovery call to build your Copilot ROI measurement framework.