Written by: Mark Hull, Co-Founder and CEO, Exceeds AI | Last updated: July 2, 2026
Key Takeaways
- Engineering leaders at 100–999 engineer companies need commit-level proof that AI coding investments deliver ROI, yet most analytics tools predate AI coding.
- DX, Jellyfish, LinearB, and Swarmia rely on surveys, metadata, or heuristics that cannot provide auditable, line-level evidence of AI impact on code quality or productivity.
- Exceeds AI writes portable, line-level attestations into the customer’s own repo at commit time, covers five major AI tools with deep fidelity, and surfaces first insights within about an hour.
- Only Exceeds AI combines provenance depth, multi-tool coverage, outcome-based pricing, and in-agent coaching to connect measurement with real developer behavior change.
- Teams ready to prove AI ROI quickly should start a free Exceeds AI pilot by connecting their repo.
Why DX Falls Short for AI-Era Analytics
DX (GetDX) centers its AI analytics on developer experience surveys and an always-on closed-source CLI daemon that captures AI usage signals and routes everything to DX Data Cloud. This design produces subjective sentiment data and proprietary metadata. Nothing lives in your own repo, nothing is auditable at the line level, and the weakest capture tier falls back to filesystem-change heuristics. When a board asks for commit-level proof that AI is improving productivity and quality, survey averages and heuristic estimates cannot meet that standard.
Six Criteria That Define AI Analytics Readiness
These six criteria form a practical checklist for leaders who need commit-level AI ROI proof. Any platform that fails on provenance depth or multi-tool coverage cannot answer the board-level questions engineering leaders now face.
1. Provenance depth. The platform must capture which specific lines are AI-generated, by which tool, in which interaction mode, instead of estimating from metadata.
2. Multi-tool coverage. It should attribute code across Cursor, Claude Code, Codex, GitHub Copilot, and Windsurf simultaneously, not just one vendor’s telemetry.
3. Time to first insight. New customers should see actionable data in hours, not wait weeks or months.
4. Portability and auditability. Attestations should live in the customer’s own repo, survive outside the vendor’s platform, and avoid lock-in to a proprietary cloud.
5. Actionability beyond dashboards. The platform needs to tell managers what to do next, not stop at descriptive charts.
6. Pricing and operational footprint. Pricing should align with outcomes instead of punishing per-seat models, and the capture layer should avoid heavy, always-on processes on developer machines.
See how your stack measures against these six criteria
1. Exceeds AI: Commit-Level AI Impact With Coaching
Exceeds AI is an AI-impact analytics platform built by former engineering executives from Meta, LinkedIn, Yahoo, and GoodRx. Its two-product architecture, the Exceeds AI platform plus Exceeds Ink as the on-machine provenance layer, connects what an engineer typed locally to outcomes in the codebase.
How provenance works. Exceeds Ink is a lightweight Rust binary that installs via standard Git hooks. It avoids long-lived daemons, PATH-shimmed git binaries, and global git config mutation. At commit finalization, per-tool checkpoint materializers for Claude Code, Cursor, and Codex resolve edit evidence against the actual working tree and write a structured attestation as a Git Note at refs/notes/exceeds-ink. Every line carries its tool, model, session, interaction mode, and timestamp. Lines that cannot be confidently attributed are recorded as unknown_lines instead of being silently rolled into “human” or “AI.” That Git Note lives in the customer’s own repo, remains portable across forks and mirrors, and is readable by any Git client without Exceeds infrastructure.
Multi-tool coverage. Exceeds AI ships five first-class adapters with deep per-tool fidelity for Claude Code, Cursor, Codex, GitHub Copilot, and Windsurf, plus lighter-weight detection across up to approximately 50 AI tools. Exceeds AI founder Mark Hull used Claude Code to develop three workflow tools totaling around 300,000 lines of code at a token cost of approximately $2,000. Exceeds Ink makes this kind of token-spend-to-output correlation measurable at the org level.
Actionability. The platform goes beyond dashboards by embedding coaching directly into the developer workflow. Coaching Surfaces and Best Practices Insights use a LangGraph-backed pipeline to identify the top three skills worth scaling across the organization. Ink-prompting-coach then delivers those patterns as a SKILL.md file and slash command inside the developer’s own Claude Code or Cursor agent, so coaching appears where the work happens, not in a separate dashboard. When a pattern proves effective, Skill Transfer distributes it org-wide, and when it does not land, Rollback retracts it cleanly.
Time to value. Teams see first insights within about 60 minutes of GitHub authorization. Complete historical analysis typically finishes within four hours, and real-time updates arrive within five minutes of new commits.
Pricing. Pricing is outcome-based, not per-contributor. The Pro plan costs $49 per manager per month (Early Partner Pricing). A free seven-day pilot covers one seat, up to ten contributors, and five repositories.
Limitations. Repo access is required for code-level analysis. Teams with strict compliance constraints that block read-only repo access should evaluate the in-SCM deployment option. The sweet spot is 50–1,000 engineers, since organizations below 50 engineers may not yet face the leadership challenges the platform targets.
Best fit. VPs of Engineering and engineering managers at 100–999 engineer companies who must prove AI ROI to a board within weeks, govern multi-tool token spend, and scale adoption without micromanaging every pull request gain the most value.
Start your free seven-day pilot with up to ten contributors
2. DX (GetDX): Experience Surveys With Heuristic AI Signals
DX (GetDX) is an engineering intelligence platform, now part of Atlassian, that combines developer experience surveys with AI Code Insights and Agent Experience modules. Its compliance posture is strong, including SOC 2, ISO 27001, ISO 27701, and Data Privacy Framework, and its Atlassian distribution gives it natural reach into Jira and Bitbucket shops.
Strengths. DX offers broad engineering-intelligence coverage, an established survey methodology for measuring developer sentiment, Atlassian ecosystem integration, and enterprise-grade compliance certifications.
Limitations. AI Code Insights relies on an always-on closed-source CLI daemon, and all attribution data lives in DX Data Cloud instead of the customer’s own repo. The weakest capture tier uses filesystem-change heuristics. By Exceeds’ own assessment, heuristic and watermark-based AI detection tops out around 20–25% accuracy. DX does not provide portable Git Notes attestations, longitudinal outcome tracking for AI technical debt, or an in-agent coaching layer. Pricing is bespoke enterprise, with a median ARR around $51,520 (Vendr-reported), and onboarding is consulting-heavy, often taking weeks to months.
Best fit. Large enterprises already standardized on Atlassian tooling that need developer sentiment measurement and compliance-grade reporting, and that do not yet require commit-level AI provenance, align well with DX.
3. Jellyfish: DevFinOps and Budget Alignment
Jellyfish is a DevFinOps platform that helps CFOs and CTOs understand engineering resource allocation by aggregating Jira and Git metadata into financial reporting views.
Strengths. Jellyfish provides strong financial alignment reporting, helps connect engineering output to budget and headcount decisions, and is recognized by enterprise procurement teams.
Limitations. Metadata-only analysis cannot distinguish AI-generated from human-authored lines. Jellyfish cannot prove whether AI investments improve or degrade code quality, cannot track AI technical debt longitudinally, and offers no multi-tool AI attribution. Setup commonly takes around nine months to show ROI, which conflicts with board questions that arrive in weeks. Pricing is opaque and per-seat.
Best fit. CTOs and CFOs at large enterprises who need engineering-to-finance reporting and are not yet asking commit-level AI ROI questions benefit most.
4. LinearB: Workflow Automation Without AI Provenance
LinearB is an engineering productivity platform focused on workflow automation, including cycle time reduction, pull request review optimization, and SDLC process metrics.
Strengths. LinearB measures what happened in the development workflow, and its WorkerB automations can reduce review latency. Initial setup is relatively fast compared to Jellyfish.
Limitations. LinearB remains metadata-only and cannot distinguish AI versus human contributions or prove AI ROI at the code level. Users have reported significant onboarding friction and, in some cases, surveillance concerns around data collection. LinearB improves the review and merge process but remains blind to the AI-assisted creation phase, which is where most AI impact and risk originate. Pricing uses a per-contributor model with a complex credit structure.
Best fit. Engineering teams focused on traditional SDLC workflow optimization that are not yet measuring AI-specific outcomes find LinearB adequate.
5. Swarmia: DORA Metrics for Smaller Teams
Swarmia is a developer productivity platform built around DORA metrics, team health tracking, and Slack-based nudges for engineering habits.
Strengths. Swarmia offers fast setup, a clean interface, and effective support for teams that want DORA metric baselines and lightweight Slack-based engagement without heavy onboarding.
Limitations. Swarmia was built for the pre-AI era and has limited AI-specific context. It tracks traditional delivery metrics but cannot attribute outcomes to specific AI tools, cannot prove AI ROI, and has no longitudinal AI technical debt tracking. It functions as a dashboard without a behavior-change or coaching layer. Pricing is per-seat.
Best fit. Smaller engineering teams under 100 engineers that want DORA metric visibility and developer engagement nudges without AI-era analytics requirements are the primary audience.
Synthesis: Why Provenance and Footprint Separate Exceeds
The central divide in this comparison is not feature breadth. The real gap is whether a platform can produce auditable, commit-level evidence that AI is improving or degrading engineering outcomes. Jellyfish, LinearB, and Swarmia were built before AI coding tools existed, so they track metadata that cannot answer the AI ROI questions boards now raise. DX has moved toward AI analytics but, as noted earlier, its heuristic approach produces sentiment data and estimates rather than the commit-level proof boards now require.
Exceeds AI is the only platform in this comparison that writes a portable, line-level, machine-readable attestation into the customer’s own repo at commit time, covers five AI tools with deep per-tool fidelity, and closes the loop back into the developer’s own AI agent with in-agent coaching. The operational footprint difference also matters at scale. No long-lived daemon, no PATH-shimmed git binary, and no global git config mutation give fleet operations and security teams a materially simpler deployment conversation.
Get commit-level AI provenance in your repo within an hour
Selection Guidance by Team Size and AI Maturity
These profiles outline common team situations and map them to the platform that best fits each context. Treat them as practical patterns rather than rigid categories.
100–999 engineers, active multi-tool AI adoption, board pressure on ROI. Exceeds AI is the direct fit. The combination of Exceeds Ink’s commit-level provenance, coverage across all five major AI coding tools, fast setup, and outcome-based pricing addresses every dimension of this scenario. With 51% of professional developers using AI tools daily, the measurement gap is already present and compounding.
Large enterprise, Atlassian-standardized, compliance-first. DX is worth evaluating alongside Exceeds AI. DX’s closed-source daemon and DX Data Cloud storage require a trust-on-faith security posture that Exceeds Ink’s code-visible capture path avoids.
Finance-reporting focus, engineering-to-budget alignment. Jellyfish serves this use case but should be paired with a code-level AI analytics layer. Jellyfish alone cannot answer AI ROI questions.
Traditional SDLC optimization, no immediate AI ROI mandate. LinearB or Swarmia cover workflow and DORA metric baselines. When AI ROI questions arrive, neither platform will answer them without a provenance layer added on top.
Teams under 50 engineers. Swarmia’s lightweight setup and DORA focus may be sufficient at this stage. Exceeds AI’s strongest value emerges at 50+ engineers, where manager span-of-control pressure and multi-tool AI chaos create the leadership challenges the platform is designed to solve.
Connect your repo and see first insights in 60 minutes
Frequently Asked Questions
What is DX (GetDX) and AI?
DX (GetDX) is an engineering intelligence platform, now owned by Atlassian, that combines developer experience surveys with AI-focused modules called AI Code Insights and Agent Experience. The AI component attempts to measure how engineers interact with AI coding tools by capturing usage signals via a closed-source CLI daemon and surfacing sentiment data about the developer experience with those tools. DX’s approach to AI analytics is experience-oriented and focuses on how developers feel about AI and on aggregate usage patterns. DX does not produce commit-level, line-level proof of AI’s impact on code quality, productivity outcomes, or technical debt, which requires a code-visible provenance layer that DX does not currently offer.
Does DX (GetDX) use AI?
DX (GetDX) incorporates AI in two ways. Its AI Code Insights and Agent Experience modules track developer usage of external AI coding tools like GitHub Copilot and others. DX (GetDX) also uses AI internally to process and surface insights from the survey and usage data it collects. However, DX’s analytics infrastructure relies on a closed-source daemon and heuristic detection methods rather than authoritative, client-level AI capture. The attestation data DX (GetDX) produces lives in DX Data Cloud, not in the customer’s own repository, and remains neither portable nor independently auditable in the way a Git Notes-based provenance layer would be.
Is DX (GetDX) still used?
DX (GetDX) remains an active product with enterprise customers, particularly in organizations already standardized on Atlassian tooling through Jira and Bitbucket. Its acquisition by Atlassian has expanded its distribution reach. DX (GetDX) holds strong compliance certifications, including SOC 2, ISO 27001, and ISO 27701, which matter to regulated enterprise buyers. DX (GetDX) faces increasing competitive pressure in the specific use case of commit-level AI ROI proof. Engineering leaders who need to answer board questions about AI spend with hard evidence, not survey averages, are finding that DX’s survey-driven, heuristic approach does not meet that bar. That gap is the problem platforms like Exceeds AI were built to solve.
Conclusion: Choose a Platform That Proves AI ROI
Engineering leaders now face a choice between measurement that produces evidence and measurement that produces estimates. Survey averages and metadata dashboards worked for the pre-AI era. In 2026, with multi-tool AI adoption accelerating and boards asking for hard proof, leaders need commit-level, line-level, tool-attributed provenance that lives in the customer’s own repo and can be audited by anyone with access.
Exceeds AI delivers that level of proof at the speed described earlier, across all five major tools simultaneously, and pairs it with prescriptive coaching that closes the loop inside the developer’s own AI agent. No long-lived daemon, no per-seat data tax, and no nine-month wait for ROI that should surface in week one.