CyberAdvisors · Team Structure & Dependencies

Team Profiles

Four Groups. Defined Roles.

Each pillar has a distinct ownership domain. The pipeline from request to deployed agent flows through all four — in sequence.

Deploy

Building Copilot agents

Power Automate flows & automation

Power Platform admin & governance

Agent deployment & publishing

Prompt gallery curation & tooling

Infrastructure & licensing readiness

Active · Tactical Execution

Relay

Evaluate AI request viability

Manage process pipeline A → B

Change Advisory Board (CAB) function

Hand off approved requests to Deploy

Iterate & refine agent intake process

Active · CAB Model

Gate

Business risk identification for AI use

API integration security evaluation

Data connection & compliance review

Protect the business from AI exposure

Evaluate what connectors & sources are safe

Hard Gate · All Approvals

⭐

Anchor

Public-facing AI value communication

See & sell AI value within departments

Champion adoption across the org

Strategic direction & budget authority

Unblock cross-group escalations

Champion · Evangelist · Unblock

Cross-Team Dependency Map

Where We Depend on Each Other

Every stage of the AI request-to-agent pipeline creates a dependency. A request doesn't go straight to Deploy to build — it flows through Relay for viability, Gate for safety, and needs Anchor visibility throughout.

Stage 01

Request & Intake

Relay → Gate Receives AI agent or use-case request; routes to Gate to assess business risk before any build begins HARD DEPENDENCY

Anchor → Relay Champions the request at departmental level; provides strategic context and ensures alignment before pipeline engagement EARLY SIGNAL

Stage 02

Viability & Risk Review

Gate → Relay Returns risk assessment — data exposure, API connection concerns, compliance flags — to Relay to determine build viability HARD DEPENDENCY

Relay cannot approve the request or hand off to Deploy without Gate clearance. This is a hard stop in the pipeline.

Stage 03

Handoff to Build

Relay → Deploy Approved request handed off with full context — use case, scoped requirements, risk constraints — so Deploy can begin building the agent or flow HARD DEPENDENCY

Anchor → Deploy Provides departmental context and visibility so the agent is built for real-world use — not in isolation COORDINATED

Stage 04

Build & Deploy

Deploy → Relay Builds and tests the agent or flow; returns to Relay with working output for process validation before wider rollout HARD DEPENDENCY

Anchor → Deploy Champions the deployed agent within their department; drives initial adoption and surfaces real-world feedback ADOPTION DRIVER

Without Anchor visibility and departmental buy-in, agents get built but not used. Adoption is owned by Anchor — not Deploy.

Stage 05

Iterate & Scale

Relay → Deploy Collects process feedback from users; refines requirements and re-engages Deploy for agent iteration or new build requests CONTINUOUS LOOP

Anchor → Anchor Brings new departmental AI ideas back to the committee — fueling the next cycle of requests through the pipeline PIPELINE FEED

How to read this diagram

HARD DEPENDENCY Work cannot move forward until this handoff is completed. If the upstream group doesn't deliver, the downstream group is fully blocked — no workaround.

COORDINATED Both groups need to work together on this step. Neither fully owns it alone — timing and communication between them matters.

REPORTING / FEED Information flows one way for visibility or continuity. Not a blocker — but skipping it creates blind spots and breaks the next cycle.

Hard stop. This callout means a stage cannot proceed at all — not slowly, not partially. Everything stalls until the named condition is met.

Responsibility Matrix

Who Does What

Every key activity mapped to a clear ownership designation. No activity should have more than one Accountable (A). Where that column is empty, ownership is undefined — and that is a gap that needs to be closed.

Activity	Deploy	Relay	Gate	Anchor
AI tool intake & evaluation Assess fit, brief preparation, SLA tracking	A	I	C	I
Security & compliance gate All risk assessment, sign-off, data classification	R	—	GATE	I
Committee approval vote Review, deliberate, approve / reject / defer	R	I	C	A
Training program design Curriculum, delivery method, schedule	C	A	—	I
Stakeholder communications Rollout comms, change narrative, resistance mgmt	I	A	—	C
Tool deployment & go-live Provisioning, onboarding, technical rollout	A	R	I	I
Adoption tracking & reporting Usage metrics, KPI dashboard, committee reports	A	C	I	I
P1 incident response Immediate action, suspension, notification	R	I	GATE	UNBLOCK
Budget allocation Tooling spend, licensing, resource requests	R	I	I	A
Policy exception approval AUP exceptions, acceptable use edge cases	R	C	C	A

A Accountable — owns the outcome

R Responsible — does the work

C Consulted — input required

I Informed — kept in the loop

GATE Gate — must sign off; blocks progress

UNBLOCK Unblock — escalation authority

Open Items · What We Need to Resolve

Setting Up for Success

Identified in our first meeting. These are the blockers and open decisions that must be resolved before the committee can operate at full capability.

Critical · Blocks Deploy

Power Platform Admin Center Not Finalized

Deploy cannot properly deploy and govern Copilot Studio agents until the Power Platform admin center is correctly configured. This includes environment setup, security settings, and compliance posture. Without it, agents are being built without a proper home to publish them from.

Critical · Blocks Org-Wide Use

No Shared Prompt Gallery Repository Yet

Every group member is curating prompts individually with no shared destination. The committee agreed a centralized prompt gallery is needed — with a quality gate process (submit → approval → publish). Until a SharePoint or Teams site is stood up, organizational knowledge stays siloed.

Warning · Licensing & Cost Risk

Agent Publishing Model & Credit Pack Strategy Undefined

Agents published beyond Teams channels (to web, SharePoint external, etc.) consume Copilot Studio credit packs — which are not yet provisioned. The committee needs to define what gets published where, who can consume it, and what the credit cost model looks like before agents go wider than internal staff.

→ Action Required · All Groups

Group Leaders Need to Schedule 30-Min Scoping Sessions

Each group leader was tasked with scheduling a 30-minute session before the next full committee meeting to define their group's specific AI focus areas, subtasks, and initial deliverables. These sessions are the foundation for the next meeting's progress report.