Frameworks Library

    Delivery · Evolved from Dual-Track Agile

    Dual-Track Agile with AI Guardrails

    Parallel discovery and delivery tracks, with explicit gates for AI-affecting work: eval readiness, guardrail implementation, and monitoring before any delivery track item can ship.

    Maturity: EstablishedReversibility: ModerateAI Risk: HighOversight: High

    Executive Guidance

    How to hold this framework as a leader

    Dual-track separates discovery from delivery so a team can both ship and learn. The framework's value is not the two tracks but the disciplined gate between them: nothing crosses into delivery without evidence, and nothing in delivery goes to customers without acceptance criteria.

    In AI-era practice this gate is the eval set. Discovery-track prototypes that touch customers without an eval spec are the most common source of preventable customer harm.

    When to Apply

    • Teams shipping continuously while validating new opportunities
    • Organizations introducing AI features into existing products
    • Environments requiring both learning velocity and shipping discipline

    AI-Era Notes

    Discovery-track AI experiments must never leak into production without crossing delivery-track gates. The evaluation set is the artifact that transitions between tracks.

    Key Trade-off

    Discovery-track AI prototypes can leak into production without crossing evaluation gates.

    Operating Sequence

    The order in which to install the framework

    1. 01Structure the team into a discovery cadence and a delivery cadence with distinct rituals.
    2. 02Define the gate between them: what must be true for an idea to enter delivery.
    3. 03For AI-affecting work, require an eval spec and guardrails before the gate opens.
    4. 04Do not allow discovery-track prototypes to reach customers outside of a scoped test.
    5. 05Review track health separately; a stalled delivery is a different problem than a stalled discovery.

    Key Artifacts

    The documents this framework produces

    Discovery Ritual Kit

    Weekly touchpoint, assumption test, and synthesis cadence.

    Owner: Product + Research

    Track Gate Definition

    Criteria that must be met to transition from discovery to delivery.

    Owner: Product Leader

    Delivery Definition of Done

    Acceptance and eval criteria for release.

    Owner: Product + Engineering

    Operating Checklist

    What "good" looks like when installed

    Discovery Track

    • Weekly customer touchpoint is calendared.
    • Prototypes with real customers are scoped, timeboxed, and consented.
    • Discovery outputs feed the assumption register.

    Delivery Track

    • Every delivery item has a PRD (and eval spec if AI-affecting).
    • Release runbook including rollback is documented.
    • Monitoring is instrumented before, not after, launch.

    Common Antipatterns

    • Discovery-track prototypes shipped to real customers without evals
    • Delivery track blocked waiting for perfect discovery

    Boardroom Questions

    • Where has a discovery prototype crossed into production without a gate?
    • Is our delivery track blocked waiting for discovery certainty that will never arrive?
    • What is the health of each track separately?

    Pairs With

    Continuous DiscoveryPRD + Eval Spec