Agentic Workflow: Delivery & Prioritisation Copilot

In large programmes, delivery friction often comes from fragmented information — updates spread across emails, pings, Jira, decks, and people’s heads.

This concept explores how a multi-agent system could support delivery teams by structuring information, highlighting risks, and supporting prioritisation — without replacing human decision-making.

Key issues

• Too many inputs, not enough synthesis
• Priority decisions rely on tribal knowledge
• Reporting is manual, repetitive, and late

Proposed agentic setup (conceptual)

Agents (clear + simple):

• Front-door Agent – summarises new requests / updates
• Dependency Agent – identifies clashes and sequencing risks
• Prioritisation Agent – applies agreed rules (risk, value, effort) to prioritize backlog
• Reporting Agent – drafts weekly status & exec summaries

<aside> 🔍

Key principle: agents recommend, humans decide.

</aside>

Example Agent flow (mermaid)

flowchart TD
		A[Email/Jira/doc] --> |1. New request arrives| B[Front-door Agent]
    B --> C{ReAct}
    C -->|2. Checks existing roadmap| D[Dependency Agent]
    C -->|3. Checks backlog| E[Prioritization Agent]
    C --> F[Reporting Agent]
    D --> C
    E --> C
    C --> |4. For Approval| G[Human]