Velocity Framework
Core Philosophy

The AI-Augmented Product & Delivery Function

How product and delivery roles use AI within the Velocity Framework.

The Velocity Framework is engineer-centric, but it only works when product and delivery operate with the same AI-native mindset. This page defines how those functions plug into the framework without prescribing a rigid org chart.

Product Work in an AI-Native World

Product roles (PM, PO, Tech Product) use AI to increase the clarity and usefulness of the input that reaches engineers.

What “good input” looks like

  • Framed around outcomes, not features
    Start from the problem, constraints, and success metrics before naming screens or APIs.
  • AI-friendly structure
    Requirements are written in a way that tools can consume directly: clear sections, explicit constraints, example payloads.
  • Tight link to the 6-step loop
    Tickets naturally lead into Step 1 (Define & Clarify) instead of forcing engineers to reverse-engineer intent.

Practical patterns

  • Use AI to turn unstructured notes into:
    • Refined problem statements.
    • Initial user journeys.
    • Draft acceptance criteria (Given/When/Then) that engineers can critique.
  • Ask AI to stress-test requirements:
    • "List edge cases for this flow. What’s missing?"
    • "Generate 5 failure scenarios that would break this solution."

The output of product work should make Step 1 (Define & Clarify) faster, not optional.

Delivery Operations in the Loop

Delivery roles (Project Manager, Delivery Lead, AI Delivery Architect) are responsible for keeping the system of work healthy.

They use AI to:

  • Maintain clean backlogs (no zombie tickets, clear priorities).
  • Generate and maintain status artifacts (weekly summaries, risk registers) from real data instead of manual retyping.
  • Monitor flow metrics and prompt teams when the system drifts (long-running PRs, stalled tickets, repeated blockers).

Examples:

  • Summarize the last 20 PRs and surface themes: "Where are we spending effort? Where are reviews slow?"
  • Turn meeting transcripts into concise, linkable decisions and action items.

Shared Responsibilities with Engineering

AI-augmented product and delivery are not separate silos; they share responsibilities with engineering:

  • Clarify before commit
    If a ticket is ambiguous, it is everyone’s problem. Product, delivery, and engineering use AI together to refine it.
  • Make artifacts first-class
    Specs, plans, and decisions are written down and updated, not left in chats. AI helps draft, but humans own correctness.
  • Respect constraints
    Security, privacy, and client-specific rules apply to all roles. Product and delivery must understand how those rules affect which tools and models can be used.

Later sections (Governance, AI Delivery Stack, Product & Delivery Workflows) go deeper into how to implement these patterns in a real pod. This page is the philosophical baseline: product and delivery are AI-augmented partners in the same loop, not external requesters handing tasks to engineering.