Written by: Mark Hull, Co-Founder and CEO, Exceeds AI | Last updated: July 9, 2026
Key Takeaways
- DX is an Atlassian-owned engineering intelligence platform that relies on surveys and metadata to track developer productivity, but it lacks the code-level visibility needed to measure AI ROI.
- Leading companies across technology, finance, healthcare, and enterprise sectors use DX, yet its architecture cannot prove which lines of code are AI-generated or track long-term quality outcomes.
- DX’s closed-source capture model and non-portable data create governance and security limitations that modern AI analytics platforms must overcome.
- Exceeds AI addresses these gaps with portable Git Notes attestation, multi-tool adapters, and longitudinal tracking that connects AI authorship to downstream quality metrics.
- See how Exceeds AI provides code-level proof of AI impact.
DX (getdx.com) as an Atlassian engineering intelligence platform
DX (getdx.com) is an engineering intelligence platform, now owned by Atlassian, that helps organizations track developer productivity through structured surveys and system metadata. Its core offering combines developer experience measurement, drawing on the SPACE and DevEx frameworks, with workflow signals pulled from Git hosts, CI/CD pipelines, and issue trackers. The platform produces dashboards of cycle times, review latency, and developer sentiment scores that highlight friction across the software delivery lifecycle.
DX customer segments and what they reveal about adoption
DX’s published customer base spans several sectors, which shows how widely survey-and-metadata analytics have spread across modern engineering organizations. Each segment reflects different maturity levels, compliance requirements, and expectations for AI-era measurement.
Technology & SaaS companies are early adopters that seek competitive advantage through developer productivity improvements:
- GitHub
- Dropbox
- Vercel
- Amplitude
- Grammarly
- Xero
- Intercom
- Lattice
Financial Services organizations operate under strict regulatory and audit expectations while modernizing legacy stacks:
- Adyen
- Vanguard
- BNY
- Block
- Brex
- Betterment
- Paysafe
Commerce & Marketplace companies balance rapid feature delivery with reliability at global scale:
- Booking.com
- Uber
- Mercari
- Vinted
- Thumbtack
Healthcare & Life Sciences organizations combine strict compliance with complex data and safety requirements:
- Pfizer
- One Medical
- Hinge Health
- Recursion
Enterprise & Consumer brands manage large, distributed engineering teams and long-lived codebases:
- Procter & Gamble
- Dell Technologies
- SiriusXM
- NIQ
The breadth of this list reflects DX’s genuine strength: broad applicability across industries and team sizes. That reach gives DX visibility into industry-wide trends, and DX’s Q1 2026 AI Impact Report found AI coding tool adoption at 93% across more than 400 companies, a figure that underscores both the platform’s market penetration and the urgency of the measurement gap it cannot close.
DX and Atlassian: implications for buyers
As of 2026, DX (getdx.com) is an Atlassian company. Atlassian acquired DX to strengthen its engineering intelligence capabilities alongside Jira and Bitbucket, giving DX access to Atlassian’s enterprise distribution network and compliance infrastructure.
For existing DX customers, the acquisition signals deeper integration with the Atlassian ecosystem. For engineering leaders evaluating platforms, it means DX’s survey-and-metadata architecture now travels with Atlassian’s enterprise sales motion, including bespoke licensing and consulting-heavy onboarding. DX’s AI Code Insights has a median ARR around $51,520 as the table-stakes commitment (per Vendr), with enterprise sales as the primary entry point.
DX vs modern AI analytics platforms in the boardroom
The companies listed above chose DX to solve a real problem: understanding developer productivity at scale. That problem remains. The question leadership must answer has changed.
In 2025 and 2026, boards no longer focus on how developers feel about their tools. They ask whether AI investments are paying off and whether leaders can prove it. DX’s architecture was not built to answer that question because it cannot trace AI adoption to business outcomes.
DX’s own longitudinal study found that as AI tool usage increased by an average of 65% across more than 400 companies, median PR throughput rose only about 8%. This gap is visible in DX’s own data and illustrates the measurement problem precisely. The gap between adoption and outcome is exactly where code-level analysis is required, and metadata cannot bridge it.
See how Exceeds AI measures that adoption-to-outcome gap with code-level proof.
DX (getdx.com) capture model and onboarding constraints
DX’s AI Code Insights and Agent Experience modules capture AI usage through an always-on, closed-source CLI daemon that transmits aggregates to DX Data Cloud. This architecture creates three compounding problems.
First, the weakest tier of its three-tier capture model relies on filesystem-change heuristics, which cannot reliably distinguish AI-generated code from human edits. Second, all attribution data lives in DX Data Cloud, and nothing is written to your own repository, so you lose provenance if the contract ends. Third, security teams must trust a closed-source binary they cannot inspect, which creates an audit gap that many compliance frameworks cannot accept. Onboarding is enterprise sales-led, and setup timelines are measured in weeks to months.
The deeper architectural problem is that DX’s provenance data is proprietary and non-portable. When a board member, auditor, or legal counsel asks for a machine-readable record of which lines in a given file were AI-generated, DX cannot produce one from your own Git history. The answer lives in DX’s cloud, not yours.
Survey-based impact measurement compounds this limitation by conflating sentiment with outcomes. A 2026 mixed-methods study at BNY Mellon, itself a DX customer, found that around 60% of developers reported less than one hour of weekly time savings from GitHub Copilot, with only a weak correlation (0.34) between satisfaction scores and reported time savings. Sentiment and productivity are not the same signal.
The quality risk remains equally invisible to metadata tools. A Carnegie Mellon study of Cursor-adopting repositories found that static analysis warnings and code complexity increased following AI adoption, with both degradations persisting long-term. A platform that cannot read diffs cannot detect this pattern until it surfaces as incidents.
Get portable, auditable provenance that lives in your own repository.
Exceeds AI architecture, coverage, and outcome tracking
Exceeds AI was built specifically for the measurement problem that DX’s architecture cannot solve. The platform’s foundation is Exceeds Ink, an on-machine provenance layer that captures AI authorship across Claude Code, Cursor, Codex, GitHub Copilot, and Windsurf with line-level fidelity. Ink writes a structured attestation as a Git Note at refs/notes/exceeds-ink, which is portable, machine-readable JSON that lives in your own repository and travels across forks and mirrors.
The architectural difference matters for three reasons. First, Ink uses a hook-direct model with no long-lived daemon, no PATH-shimmed git binary, and no global git config mutation. Because it fires on standard Git hooks and exits immediately, your CISO can audit the entire capture code in an afternoon, with no persistent process to monitor or background service to secure.
Second, attribution is deterministic at commit finalization, not asynchronous after the fact, so a fast git push cannot race ahead of the attestation. Third, the Git Note is yours. It survives outside the Exceeds platform, is readable by any Git client, and answers governance questions with evidence rather than estimates.
On the multi-tool question, engineers in 2026 rarely rely on a single AI coding tool. Elite engineering teams in 2026 achieve AI-assisted code shares of roughly 60–80%, exemplified by Spotify at 73% of pull requests. These teams draw on Cursor for feature work, Claude Code for large-scale refactoring, Codex for batch transforms, and GitHub Copilot for inline autocomplete. Exceeds Ink provides five first-class adapters with deep per-tool checkpoint materializers for Claude Code, Cursor, and Codex, plus lighter-weight detection across up to approximately 50 AI tools. DX’s capture model goes dark when engineers switch tools outside its telemetry coverage.
On outcomes, Exceeds tracks AI-touched code longitudinally over 30 or more days. It monitors incident rates, rework patterns, and maintainability issues that only surface after initial review. Faros’s 2026 AI Engineering Report, analyzing telemetry from 22,000 developers across 4,000 teams, found code churn up 861% and the incident-to-PR ratio more than tripled under high AI adoption. Detecting that pattern requires reading diffs over time, not surveying developers about their experience.
On prescriptive guidance, Exceeds moves beyond dashboards. Coaching Surfaces and the ink-prompting-coach skill distribute actionable guidance directly into the developer’s own Claude Code or Cursor agent. When one team finds a pattern that works, Skill Transfer and Rollback lets managers distribute it as a versioned skill across the organization. Managers report saving 3–5 hours per week on performance analysis and productivity questions, time recovered from staring at descriptive metrics that offer no next action.
On pricing and setup, Exceeds uses outcome-based pricing with no per-contributor data tax. First insights are available within 60 minutes of GitHub authorization. Complete historical analysis finishes within four hours. That timeline contrasts with DX’s enterprise sales gate and weeks-to-months onboarding.
Start a seven-day pilot and see code-level AI impact within hours.
Frequently Asked Questions
How does DX compare to code-level AI provenance tools?
DX measures developer experience and AI usage through surveys, metadata, and a closed-source CLI daemon that sends aggregates to DX Data Cloud. It cannot distinguish which specific lines in a pull request were written by an AI tool versus a human engineer, and it cannot track the long-term outcomes of AI-touched code.
Code-level provenance tools like Exceeds AI analyze actual diffs at the commit and PR level, write a portable attestation alongside every commit, and correlate AI authorship with downstream quality metrics such as incident rates and rework patterns. The practical difference is the difference between knowing that developers feel more productive and being able to prove to a board or auditor exactly which lines were AI-generated, by which tool, and what happened to them 30 days later.
Get that proof layer in place before your next board review.
Why do metadata platforms miss AI technical debt?
As discussed earlier, metadata platforms track volume signals that AI tools inflate, such as PR cycle time, commit volume, and review latency, without measuring delivered value. A team using agentic AI tools can double its PR volume without doubling its delivered value, and metadata tools will report the volume increase as a productivity gain.
AI technical debt accumulates at the code level through subtle architectural misalignments, duplicated logic, and complexity increases that pass initial review but degrade maintainability over weeks and months. Detecting these patterns requires reading diffs over time and correlating AI authorship with follow-on edits, test coverage changes, and incident rates. No metadata platform has the inputs to perform that analysis.
Add code-level AI debt detection alongside your existing DX dashboards.
Can Exceeds AI be used alongside DX?
Yes. Exceeds AI is not a replacement for developer experience surveys or metadata dashboards. It is the AI intelligence layer that answers the questions those tools cannot answer, such as which code is AI-generated, which tool produced it, whether AI-touched code is performing better or worse than human-authored code, and what managers should do next to scale effective adoption.
Organizations that already use DX for developer sentiment measurement can add Exceeds AI to gain the code-level proof layer that DX’s architecture does not provide.
Layer Exceeds AI on top of DX for code-level AI visibility.
What makes Exceeds Ink’s provenance portable in a way DX’s is not?
Exceeds Ink writes its attestation as a Git Note at refs/notes/exceeds-ink, which is structured JSON that lives inside your own repository and travels with it across forks, mirrors, and CI systems. Any Git client can read it, and it survives outside the Exceeds platform entirely.
DX’s AI Code Insights stores all attribution data in DX Data Cloud. If your DX contract ends, that provenance data does not come with you. For governance, audit, and legal use cases, such as proving to a patent examiner what role AI played in a given file or tracing an incident back to a specific AI session, portability is not a nice-to-have. It is the requirement.
Get provenance that stays with your code, not your vendor.
How quickly can an engineering team get value from Exceeds AI?
Setup requires GitHub, GitLab, or Azure DevOps authorization plus a lightweight per-machine Ink install. First insights are visible within 60 minutes. Complete historical analysis across a repository finishes within four hours, and real-time updates appear within five minutes of new commits.
The Pro plan starts at $49 per manager per month with no per-contributor data tax. A free seven-day pilot is available with no enterprise sales gate required.
Start your free seven-day pilot today.
Conclusion: choosing AI measurement lenses for 2026
The DX customer list above represents organizations that made a reasonable choice for the pre-AI era: measure developer experience, track metadata, and surface friction. That choice remains defensible for teams whose primary question is how developers feel about their workflow.
The question most engineering leaders face in 2026 is different. Many organizations cite lack of clarity on AI’s ROI as a top challenge, and boards are not satisfied with sentiment scores. Organizations that measure AI impact across multiple dimensions often achieve greater returns than teams that measure only activity. Metadata tools measure activity. Code-level tools measure outcomes.
Three lenses clarify the decision. First, provenance: can the platform produce a machine-readable, auditable record of AI authorship that lives in your own repository? Second, multi-tool coverage: does the platform have authoritative capture across Claude Code, Cursor, Codex, Copilot, and Windsurf, or does it go dark when engineers switch tools? Third, longitudinal quality: does the platform track what happens to AI-touched code 30, 60, and 90 days after merge, or does it only report on the moment of submission?
DX answers none of these three questions. Exceeds AI was built to answer all of them with commit and PR-level fidelity, portable Git Notes attestation, five first-class AI tool adapters, and prescriptive coaching that turns measurement into adoption. It is a platform designed for the AI coding era.