Prove how your software was built
Every regulated software team can show it followed the process. Almost none can show whether a person or an agent did the work. Bloomfilter reads the record you already keep in Jira, GitHub, and CI and shows how the work actually ran: which steps a person did, which an agent did, and where a step was skipped.
What changes
The rules barely change when agents write the code. What it takes to prove you followed them does. A person who writes a requirement and links a test leaves evidence behind as a side effect. An agent that does the same work in seconds leaves a gap, unless something is watching the process.
See how the work actually ran
Bloomfilter mines the process from the systems your teams already use: tickets in Jira or Azure DevOps, commits and pull requests in GitHub, builds and tests in CI. It rebuilds each team's real process step by step and shows where work went around a review, a test, or a sign-off.
Know whether a person or an agent did each step
Agent Miner ties each agent session to the work item it touched, the pull request it produced, and the person who reviewed it. For any change you can show who or what wrote it, against which work item, and what happened next: accepted, reworked, or reverted.
Keep the evidence while the work happens
Once a policy is defined, Bloomfilter checks work against it as it moves and keeps the record. The trace exists before an assessor, auditor, or examiner asks for it, so the next audit is a report to run rather than a month of rebuilding links by hand.
- Finance
SOX change management
Separation of duties breaks quietly when the engineer who prompts the agent also approves its pull request.
- Automotive
ISO 26262 and Automotive SPICE®
An agent can produce a perfect traceability matrix in one session, so assessors are turning from the artifacts to the process that made them.
- Finance
DORA
Technology neutral cuts both ways. Agent changes get no exemption from your ICT change procedure, and no special lane through it.
- Medical
IEC 62304 and FDA
The FDA's AI guidance is about AI inside the device. The agent that writes the device's code is a question for your quality system.
- Aerospace
DO-178C and DO-330
No regulator has qualified a generative code generator, so every agent change rests on verification you can prove happened.
- Governance
EU AI Act
For most software teams it governs what you ship, not the coding assistant you use to build it.
To improve a process, you first have to understand it. With Bloomfilter, it's like an X-ray. We can finally see what is really happening, based on facts and figures, not opinions.
- Process events
- 133,729
- Rebuilt from one project's record, with V-Model traceability
- Of defects were regressions
- 72%
- Rework the BI tools could count but not explain
- Of stories skipped code review
- 18%
- The skip rate tracked the regression rate
- Of bug reports were waste at intake
- 41%
- Most were later cancelled
Source: Bloomfilter engagement with a global tier-one automotive supplier, from its ALM and version control process record, as published in “What changes for Automotive SPICE® assessors when AI agents do the development work.”
Claude Code
Cursor
GitHub Copilot
Codex
Devin
Windsurf
Gemini CLI
Amazon Q Developer
Qualtrics
Google Analytics
Jira Product Discovery
Confluence
Figma
Lucidchart
Aha!
Jira
Linear
Azure DevOps
VS Code
GitHub
GitLab
Docker
Selenium
Cypress
Postman
SonarQube
Bitbucket
Gerrit
GitHub Actions
Jenkins
CircleCI
Argo CD
Kubernetes
Datadog
Prometheus
Grafana
PagerDuty
ServiceNow
Agent share by step
How much of each process step an agent did, and how much a person did
Skipped steps
Work that went around a required review, test, or sign-off
Unreviewed agent changes
Agent changes that shipped without a person looking at them
Traceability gaps
Work items with broken or missing links to tests and verification
Time to sign-off
How long agent work waits for the person who has to approve it
Rework after review
Work that came back after it was approved, and the step it came back from
| Bloomfilter | ALM and QMS tools | GRC tools | Engineering metrics tools | Manual audit prep | |
|---|---|---|---|---|---|
| Agent or person, step by step | Yes | No | No | No | No |
| Process as it ran, across systems | Yes | No | No | No | No |
| Skipped steps found from the record | Yes | No | No | No | No |
| Trace from work item to release | Yes | Yes | No | No | Yes |
| Human sign-off recorded | Yes | Yes | Yes | No | Yes |
| Evidence kept without manual entry | Yes | No | No | Yes | No |