Skip to main content
Skip to main content
Governance & Trust

Continuous Improvement Policy

How we turn feedback, incidents and test results into concrete changes to the framework — and how you can see that happening.

Version 1.0
Last updated View version history

Maintained by Automated Governance Systems. Last reviewed August 2026. This page is app-owner maintained content, not an independent certification or audit report.

Our commitment

The framework is expected to get measurably better over time, not merely to keep working. We run a plan–do–check–act cycle: gather evidence, decide the change, ship it behind our automated checks, then confirm the measure moved.

The evidence we use

  • Support tickets and complaints, grouped by theme rather than handled one by one.
  • "This helped" and "this didn't help" feedback on knowledge base answers, which reweights future results.
  • Reported AI failures — wrong, biased or unsafe answers — treated as defects.
  • Automated accessibility scans against our published baseline, plus scheduled reports.
  • Dependency vulnerability scans, security findings and incident reviews.
  • Usage and engagement signals for the learning and governance tools, used in aggregate.

What we measure

  • Time to acknowledge and time to resolve support tickets against our published targets.
  • Repeat-issue rate: how often the same root cause returns.
  • Accessibility violations, which are expected to trend to zero and never regress past the baseline.
  • Open security findings and the age of unpatched dependencies.
  • Learner progress and mastery signals in SAMU and TRAC, used to improve content rather than to rank people.

How improvements are made

Changes are reviewed, tested and released through our pipeline, which blocks a release on failing type checks, unit tests, end-to-end tests, accessibility thresholds or dependency alerts.

Content, curriculum and guidance changes are checked against the primary source before publication, and UK English usage is enforced automatically.

Review cycle

  • Weekly: accessibility reports and support themes.
  • Monthly: security findings, dependency posture and open improvement actions.
  • Quarterly: policy review, including this page, and a review of AI behaviour against our Ethical AI Policy.
  • Annually: a full review of the framework's governance controls.

Your part in it

Suggestions are welcome and are logged in the same queue as faults. Ask HELIX if you need help, raise an improvement through the Support & escalation page, or email hello@automatedgovernancesystems.com. We will tell you what we decided and why.

Version history

Current: v1.0

Last updated . Earlier versions are listed for reference; the newest version is the one that applies.

  1. Version 1.0
    Current

    First published improvement cycle.

Questions about this policy?

HELIX — Help Assistant

Ask anything about ALFI + EDIE, governance, or the SAMU apps.

Try HELIX now

Type a question below, or pick a quick prompt to get started.

Quick prompts

HELIX is an AI assistant. Replies may be incomplete or wrong — check anything important with a person.

Try the demo