AI impact

The AI impact report helps you measure and report on the impact of AI code assistants. The report includes views for cohort comparisons, before-and-after changes, adoption trends, reported time and dollar savings, and lifecycle performance.

Data requirements

AI impact requires an attribute that groups contributors by AI usage. DX uses automatic AI usage attributes when available and falls back to the self-reported AI code assistant usage attribute.

Additional requirements by tab:

  • Before vs after and Trend correlations: Require AI adoption dates.
  • Time savings: Requires a completed Snapshot with the AI time savings workflow question enabled.
  • Dollar impact: Requires AI Code Insights, a source code management connection with PR commit data, and AI time-savings responses from a snapshot.
  • Lifecycle impact: Requires Jira or Linear and a supported source code management connection. Deployment data is optional.

Group comparisons

The Group comparisons tab compares productivity metrics between the selected Baseline and Comparison cohorts. Use it to identify differences associated with AI usage levels across your organization.

Metric differences can have causes unrelated to AI usage. Use the chart tooltips to understand each metric, and investigate other factors before drawing conclusions.

Before vs after

The Before vs after tab compares contributors’ productivity metrics before and after they reach a specified AI usage level. The comparison uses up to 90 days on each side and requires comparison windows spanning at least five days. Use it to understand how productivity and output change as individual AI usage increases.

Trend correlations

The Trend correlations tab displays changes in AI adoption alongside productivity metrics over time. Use it to identify patterns such as PR throughput increasing with AI adoption. The lines are a visual comparison, not a statistical correlation or evidence that AI caused the change.

Time savings

The Time savings tab shows annualized time savings and potential annualized savings based on Snapshot responses. Depending on your configuration and available data, it also displays dollar savings and industry benchmarks.

DX uses one of two calculations based on the Snapshot publication date:

DX calculates metrics from the full set of responses without sampling.

  1. DX multiplies each response by 48 weeks to compute annualized hours saved. The sum equals the total annualized time savings for the organization.
  2. DX divides the total annualized time savings by the number of developers to calculate the average annualized time savings per developer.
  3. DX multiplies the total annualized time savings by the fully loaded FTE cost, then divides by 2,080 work hours per year to calculate the total annualized dollar savings for the organization.
  4. DX compares current cost savings with a scenario where every developer attains the same time savings as the 75th percentile of the active cohort to calculate potential cost savings.

DX samples responses and extrapolates metrics from that data.

  1. DX calculates the average weekly time savings per developer for each cohort, such as Daily active user or Weekly active user.
  2. For each cohort, DX multiplies the average weekly time savings by the number of developers in the cohort and by 48 weeks to compute annualized time savings.
  3. DX adds the annualized time savings for all cohorts to calculate the total annualized time savings for the organization.
  4. DX divides the total annualized time savings by the number of developers to calculate the average annualized time savings per developer.
  5. DX multiplies the total annualized time savings by the fully loaded FTE cost, then divides by 2,080 work hours per year to calculate the total annualized dollar savings for the organization.
  6. DX compares current cost savings with a scenario where every developer attains the same time savings as the 75th percentile of the active cohort to calculate potential cost savings.

Dollar impact

The Dollar impact tab combines efficiency gains, agent output, oversight costs, and AI tool spend into a net dollar figure for the selected timeframe. Read AI dollar impact for the complete calculation and configuration details.

Lifecycle impact

The Lifecycle impact tab compares delivery time between the selected Baseline and Comparison cohorts. Use it to identify phases where different AI usage levels are associated with faster delivery or additional friction.

The tab shows average total lifecycle time and four phases:

  • Refinement is the time between issue creation and the first PR commit.
  • Work is the time between the first and last PR commits.
  • Review is the time between the first PR review and the last merged PR.
  • Deployment is the time between the last merged PR and the last deployment. This phase appears only when deployment data is available.

Total lifecycle time runs from the earlier of issue creation or the first commit to the later of the last merge or last deployment. Phase averages can overlap, so total lifecycle time is not the sum of the four phase averages. DX excludes non-positive phase durations and includes issues created during the selected timeframe that have at least one measured phase.

Percentage changes compare Comparison with Baseline. A lower comparison value indicates faster delivery, while a higher value indicates slower delivery. The phase tabs pair lifecycle duration with related system and Snapshot metrics when those signals are available.

The breakdown shows where lifecycle changes differ across teams, groups, or attributes. Use Lifecycle by phase when you need to inspect individual issues rather than compare AI usage cohorts.