CTTP Phase 2 - Progress Summary Dashboard

Programme Overview

All pilot teams continue to perform well. They are balancing key pilot requirements and BAU delivery.
The programme continues to evolve and mature. Many lessons are being learnt across all Service Lines, which are extremely valuable. A recommendation to review.
Risks also continue to evolve and identified. These should also be reviewed.
The scope of the pilot programme is broad to ensure all requirements and ways of working are tested. Multiple agents and droids are being built, tooling is vast, all roles are being reviewed, hundreds of process flows have been built.
Governance meetings continue to provide value and stimulate creativity, although attendance continues to vary across Service Lines and SPOCs. AI Programme Manager, Performance Excellence and Assurance continue to be the 'mainstays'.
EY-Parthenon update is to be provided following key stakeholders return from annual leave.

5

Service Lines

12

Products

37

AI Tools

30+

Agents Defined

5-14

POD Size Range

20-30%

Targeted Savings

Assurance

Risk Radar (Greenfield) • Helix Spotfire Refactoring

Two products in pilot scope. Multiple models and agents being tested; PDLC Agents, AI Orchestration, Kanban, Assurance Shared Services Model
Two AI models expected to be produced from Assurance pilot. Assurance predicts three AI models required in total. An additional Canvas model to be built and piloted, following current pilot
Target Savings: TBC

Consulting

Marketing.AI • EYWP • EY.AI for Risk-IA

Three products in pilot scope
One unified process for the CT Consulting Service Line (for all Products)
Target Savings: 20-30%

Fabric

EY Fabric Next

Fabric-Portal (brownfield) dropped from Pilot scope due to Fabric business requirements
One product remains in scope; Fabric-Next (greenfield). Fabric-Next is the evolution of the fabric experience
All products would adopt the Fabric-Next approach (AI pods + Kanban)
Risk exists as AI solution built and tested on greenfield product. Fabric expect to add Developer Workflow (brownfield) to pilot scope
Testing Aha! and Github Copilot in place of ADO
Target Savings: 20-30%

Tax

EYXP • ITTS • EYMP • Payroll • MDR

Five products in scope of pilot. MDR added
Two AI experiences being built: Local user experience (ME), and Remote (WE) experience, for team productivity
One unified process for the CT Tax Service Line (for all Products). However, teams given additional autonomy of selecting AI workflows (agents)
Target Savings: 20-30%

EY Parthenon

Competitive Edge

One product in scope: Competitive Edge
AI-orchestrated, human-governed delivery system designed to eliminate rework loops
Definition of Ready (DoR) as a hard control gate - work cannot enter delivery unless it meets DoR
Single cross-functional product pod with end-to-end ownership
Key Principle: Fix process before AI - AI amplifies existing dysfunction
Target Savings: TBC

Status by Service Line

Assurance

Risk Radar (Greenfield) • Helix Spotfire Refactoring

Progress Area	Risk Radar (Greenfield Replay) Simulated full release • 10.5 FTE → 6.8 FTE • 9 months → 3 months	Helix Spotfire Refactoring Spotfire to Power BI migration • ~5 team members • 12 months → ~1 hour per migration
Future-state process	Approach: Working through PDLC Phases (Ideation, Build, User Acceptance, Release, Maintenance). For each activity: question feature need, validate value, design AI-native activity including tools and roles, then test. Status: Ideation Phase redesign complete and under testing. Build phase redesign in progress.	Approach: Reengineering-focused, AI-native migration from Spotfire to Power BI, centered on reverse engineering existing reports and producing implementation-ready specifications. Assurance are now the auditors for Tibco (who provide Spotfire), so Spotfire cannot be used. Must reverse engineer the Spotfire analyzers to determine the functional requirements.
Agents	Agents defined for each PDLC phase with AI Potential and AI Powered Process details. Both products taking Agentic approach with human-in-the-loop controls at validation, publishing, and prioritization decision points.	Defined Agent Functions: Requirement elaboration Backlog prioritization QA/Test generation CI/CD pipeline automation Deployment simulation Dependency risk management General orchestration via Microsoft Agentic Framework Human-in-the-loop retained at validation, publishing, and prioritization decision points.
Future Roles & POD Redesign	Role titles remain the same Pod composition changing - some roles moving to shared service vs. dedicated Potential move to Kanban model Retain PI Planning, Retrospectives, System Demos Eliminate: Sprint Planning Repurpose: Daily standup, Demo, Backlog refinement Team size target: 10.5 FTE → 6.8 FTE (35% reduction)	Work framed as spec-driven AI-native migration (not net-new traditional PDLC build) Team responsibilities shifting from manual execution to agent orchestration + review/approval Human reviewers remain accountable for semantic accuracy, business-logic decisions, final publishing approvals Moderate-autonomy stages include explicit reviewer checkpoints
Tooling Plan	Tools: ASU Studio (user-friendly Factory.ai wrapper), Factory.ai, M365 Copilot, EYQ, GitHub Copilot Integrations: Azure DevOps (backlog), GitHub (code), Playwright (test automation via agents)	Tools: Spotfire bundle extraction scripts, Agentic orchestration workflows, Knowledge graph for context enrichment, MCP + LLM semantic model generation, ADK and BPA validation framework, Power BI transformation/report generation Integrations: Microsoft Agentic Framework for skill orchestration, CI/CD and automated status reporting pipeline
Early Risks	Token cost vs. efficiency gain may be cost prohibitive depending on location/quality expectation Semantic accuracy and overfitting risk in automated processes Alignment across teams on process, tools, source of truth, and handoffs Lack of clarity to individuals due to activity overlap/segregation of duties leading to duplicative work	Critical April release dates must not be put at risk Resourcing constraints for documentation efforts Semantic accuracy and overfitting risk in automated enrichment/mapping False positives/context misses in automated validation loops Data asset readiness risk remains (current slight delay)
Lessons Learned	Process redesign: Testing of AI-native process redesign is critical as we uncover new problems. Example - UX Concept: Initial flow had no standard way to share working prototype. Updated: PM builds prototype → imported into Figma → XD Lead reviews/updates → Tech Lead validates → sandbox → business testing. Token cost: Need well-established baseline for monthly token estimate by role. Token cost/coverage agreement with suppliers needs consideration in contracts/SOWs.	High-autonomy extraction and transformation steps deliver immediate efficiency gains Moderate-autonomy stages (semantic accuracy, nuanced business logic) require explicit human-in-the-loop controls Agent orchestration reduces sequential handoffs and shortens validation cycles Progressive context enrichment improves downstream quality and rework rates

Consulting

Marketing.AI • EYWP • EY.AI for Risk-IA

Unified Process: One process for all CT Consulting products. 3 products in pilot scope with 3 POD complexity tiers (Simple 5 / Medium 8 / Complex 14).

Progress Area	Status & Details (All Products)
Future-state process	One unified process for the CT Consulting Service Line (for all Products). AI tools (Factory.ai) are a core consideration. Close monitoring of token consumption required. Build vs buy tooling decisions complete.
Agents (8 Droids)	8 Droids Defined: Feature/User Story Creator Code Generator Code Reviewer Test Case Generator Test Script Generator OSS/Aquasec Finding Fixer UX Reviewer Product Document Generator
Future Roles & POD Redesign	8 Future Roles Defined: Business Domain Expert Business processes, client pain points, translate goals to requirements Prompt Engineer - COE Prompt engineering for AI tools, iterative refinement, NLP concepts AI Engineer AI-powered dev environments (Replit, Cursor), CI/CD, agentic frameworks UX/UI Designer - Curator COE UX templates, design systems, user journeys, accessibility standards Solution Architect - Curator COE Scalable architecture, cloud platforms, security/compliance, microservices QA Automation Specialist Automated testing pipelines, Selenium/Cypress, security scanning, CI/CD integration Project Lead Agile/sprint planning, stakeholder communication, risk management Knowledge Engineer ML/AI data architecture, deep learning, domain expertise, process knowledge POD Structures (FY27 H1): Simple POD 5 1-3 months (scope timelines) 20-30 story pts/sprint Straightforward, limited scope Medium POD 8 3-6 months (scope timelines) 40-60 story pts/sprint Multiple components Complex POD 14 6-12+ months (scope timelines) 70+ story pts/sprint Large-scale, high complexity Greenfield vs Brownfield differentiation established. POD restructuring for FY27 H1 underway.
Tooling Plan	10 Tools: Figma, Motif, Experience Studio (in-house), Replit, AHA, ADO + Skadoosh MCP, GitHub, Factory.ai (Droid), Cursor, GitHub Copilot CLI
Early Risks	Inadequate role-based (including seniority) usage controls driving higher tool costs Lack of real-time monitoring of token consumption (product/eng code/user/consumed tokens) Management of token limits at product level including notifications Defining the right strategy in each phase of PDLC on how to use the tools Delivery continuity at risk if budget coverage cannot be confirmed Urgent need for competency-based training and learning plans ⚠️ Critical Governance Notice: Effective governance of AI tool usage is now a delivery-critical control. Risk from insufficient guardrails and limited visibility into token consumption can lead to unplanned cost escalation. Direct threat to delivery continuity.
Lessons Learned	1. Use the right tool: GitHub Copilot: Best for inline code suggestions MS Copilot: Best for general research, summarization Factory AI: Reserve for full workflow orchestration only Not every task needs Droid - use lighter tools first 2. Choose the right model: Use /model in Droid to select based on task complexity (Simple: 0.2-0.4 tokens, Medium: up to 0.8 tokens, Complex: 1-2 tokens) 3. Optimize your prompts: State your goal clearly in the first prompt Mention only the relevant function/class, not the whole file Use #file mentions instead of pasting code Start a new chat for each new task Provide a schema or 2-3 row sample instead of full data dumps

Fabric

EY Fabric Next

AI-First Transformation: One product in pilot scope (Fabric-Next greenfield). Fabric-Portal (brownfield) dropped due to business requirements. All products would adopt the Fabric-Next approach (AI pods + Kanban). Target Savings: 20-30%

Progress Area	Status & Details
Future-state process	The EY Fabric future-state delivery process has been formally redesigned as part of the AI-First Fabric Transformation. The redesigned PDLC replaces sequential hand-offs with AI-assisted, parallel execution across ideation, build, test, and release. Operating Model - AI-First by Default: Move from a hybrid model to AI-first delivery as the standard way of working Redesign teams around smaller, senior-led, outcome-accountable PODs, augmented by AI Embed AI agents and automation into end-to-end product delivery, not isolated tasks Measure success through outcomes, flow, quality, and cost efficiency - not activity
Agents	EY Fabric has defined a clear AI-agent model supporting the future-state process: Feature and story generation Solution design assistance Code review Test case generation QA automation Human-in-the-loop checkpoints are explicitly retained for approvals, quality gates, and compliance.
Future Roles & POD Redesign	Target Team Composition (AI-First POD) Smaller, senior-led pods (5-6 people, hard cap) Cross-functional and outcome-owned, with no role handoffs AI embedded into day-to-day delivery, not treated as a specialist add-on Typical POD Structure Size: 5-6 (hard cap) 1 Product/Outcome Lead 3-4 AI-augmented Engineers 0.5 Platform/SRE (shared) Quality, DevOps, architecture embedded - not standalone Ways of Working (Kanban) Continuous flow replaces sprints Work pulled based on capacity/WIP Cycle time & throughput metrics End-to-end outcome accountability How Roles Change Product/Outcome Lead Backlog → Outcome accountability Engineers Build/test/deploy/operate; AI by default QA/Test No longer standalone; embedded automation-first DevOps/Platform Golden paths, automation, self-service
Tooling Plan	7 Approved AI Tools for Fabric: Factory.ai AI-powered dev accelerator for code gen, refactoring, documentation, tests DROID CLI Developer automation agent for setup and workflow execution GitHub & GitHub Actions Version control with CI/CD automation, builds, tests, security scans SL Assist AI-driven SDLC orchestration for requirements, UI, code, testing, docs Figma Collaborative UI/UX design and prototyping Aha! Roadmapping, ideas intake, prioritization, release planning Motif Micro-frontend UI framework with design system components GitHub Copilot AI code assistance for inline suggestions Note: Testing Aha! and Github Copilot in place of ADO
Early Risks	Adoption challenges if teams are not sufficiently socialized on the AI-enabled model Dependency on Fabric Portal onboarding for full metrics visibility Risk of over-using heavy AI agents where lighter tools would be more effective Risk exists as AI solution built and tested on greenfield product only - Fabric expect to add Developer Workflow (brownfield) to pilot scope
Lessons Learned	Early Socialization: Importance of early socialization with delivery teams before introducing AI-enabled model Remove Blockers First: Remove non-AI process blockers before introducing AI - existing inefficiencies compound with automation Real Execution Testing: Validate AI-native processes through real execution rather than design alone Selective AI Usage: Selectively use AI agents to balance efficiency and cost - not every task needs heavy AI Measure Outcomes: Success measured by value delivered, not volume of features - focus on cycle time, throughput, delivery health

Tax

EYXP • ITTS • EYMP • Payroll • MDR

Unified Process: One process for all CT Tax products. 5 products in pilot scope with POD sizing by program complexity (Small/Medium/Large) and project type (Greenfield/Brownfield). Two AI experiences: Local (ME) and Remote (WE) for team productivity.

Progress Area	Status & Details (All Products)
Future-state process	"Way We Work" (WWW) transformation program establishing versioned, continuously evolving workflows. Focus on optimal workflows where autonomous and semi-autonomous agents collaborate seamlessly with human contributors to accelerate, standardize, and improve software delivery. Key Principle: Teams given additional autonomy of selecting AI workflows (agents) while following the unified process.
Agents	Agents are evolving as new concepts such as plugins and missions become mainstay. Agents will become primitives within these constructs, leveraging skills, tools and slash commands to be used within Workflows/Missions and packaged as plugins. 22 Agents Defined: Droid Orchestrator Droid Analyst Droid Designer Droid Architect Droid Developer Droid QA Droid DevOps Droid SRE Droid Security Droid Data Droid Docs Droid Reviewer Droid Refactor Droid Test Droid Debug Droid Migrate Droid Explain Droid Learn Droid Plan Droid Research Droid Support Droid Automate Agent behavior is explicitly designed, governed, and mapped to SDLC phases.
Future Roles & POD Redesign	POD Sizing by Program Type: POD sizes differentiated by program complexity and project type (Greenfield vs Brownfield). Small Program Greenfield: 5 resources Brownfield: 7 resources Focused, single-feature delivery Medium Program Greenfield: 7 resources Brownfield: 9 resources Standard delivery configuration Large Program Brownfield Only: 12 resources Large greenfield not anticipated for FY27
Tooling Plan	7 Tools: Factory.AI AI-powered development and workflow orchestration GitHub & GitHub Copilot Version control with AI code assistance Azure DevOps Backlog management and CI/CD pipelines M365 Copilot Productivity and documentation assistance Figma & Figma Make/Replit Design and rapid prototyping Aha! Roadmapping and product planning Azure Copilot Cloud infrastructure assistance
Early Risks	Unpredictable tool/token costs Lack of continuous upskilling of resources Continuously addressing cross-cutting concerns: Cost, Security, Performance, Reliability and Quality
Lessons Learned	7 Key Lessons from Way We Work Initiative: Tools Alone Do Not Change Outcomes — Workflows Do Agentic Delivery Requires Explicit Role and Responsibility Design Cost and Token Economics Become Material Very Quickly Workflow Coverage Is More Effective Than Full SDLC Modeling Upskilling Is a Prerequisite, Not a By-Product Measurement Must Shift From Activity to Outcomes "Way We Work" Must Be Versioned and Continuously Evolving

EY Parthenon

Competitive Edge

AI-Orchestrated Delivery: Future-state PDLC is an AI-orchestrated, human-governed delivery system. Definition of Ready (DoR) as hard control gate. Single cross-functional pod owns delivery end-to-end.

Progress Area	Status & Details
Future-state process	Core Design Principles: Enforce Quality at Entry: Work cannot enter delivery unless it meets Definition of Ready - prevents defective inputs, eliminates rework loops at source AI Executes, Humans Decide: AI generates artefacts, validates completeness, executes tasks. Humans define intent, approve gates, handle exceptions Single Product Pod Ownership: One pod owns delivery end-to-end - no internal handoffs between roles Shared Services on Exception: No manual handoffs - automated triggers engage services. Human intervention only for risk, compliance, failure scenarios Automation First, Governance Always: Automation drives flow; governance enforced through control gates, auditability, human sign-off points 8-Step End-to-End Process: 1. Intake & Problem Structuring 2. Backlog Generation 3. DoR Enforcement (Critical) 4. Planning & Sequencing 5. Build & Test 6. Pre-Release Validation 7. Release & Deployment 8. Stakeholder Communication
Agents	Key Principle: Agents are aligned to control points and flow orchestration (not standalone tools). They form an orchestrated control system across PDLC. Core Agents (Phase 2 Focus): Upstream (Highest Priority) Requirement Intake Agent → PRD generation PRD/Story Generator Agent → Features + Stories + AC DoR Validation Agent (CRITICAL) → enforces readiness gate Backlog & Flow ADO Health Agent → detects missing AC/DoD/dependencies Planning & Sequencing Agent → orders backlog, identifies blockers Build & Test Code Generation Agent (Factory.ai) QA Automation Agent → test case + script generation Cross-cutting Stakeholder Communication Agent → status updates, risk alerts, decision prompts
Future Roles & POD Redesign	Operating Model Shift: From role-based delivery → responsibility-based pods with AI execution Key Shift: Humans define and validate; AI generates and executes Single Integrated Pod (Per Product) End-to-end ownership from intake → UAT Business Product Manager → value & intent Architect → feasibility & design authority Full-stack Engineer → build + validation UX Designer → experience design What Changes BA: Strengthened to capture business knowledge Scrum Master: Reduced - functions replaced by agents Architect: Moves closer to code, validates AI output Engineers: Forward Deployment Engineers, closer to business QA: Shift from manual → validation + edge cases PMs: Focus on governance, CAB, budget, compliance Shared Services Layer: InfoSec/DevSecOps, Release/CAB, Platform/DevOps, QA Test Strategy, Architecture Governance, Data/AI Platform Support
Tooling Plan	Key Principle: Small number of core tools + orchestrated agents 6 Core Tools: Factory.ai Agent creation & orchestration ADO-integrated Agents Backlog + governance GitHub Copilot Developer assist M365 Copilot / ChatGPT / Claude Reasoning + generation Figma UX artefacts GitHub Code repository Integration Points: Aha!/Intake → ADO, ADO → Factory.ai agents, Figma ↔ ADO (UX linkage), Pipeline → QA/DevOps
Early Risks	Key Risk Statement: Without strict control gates and governance, AI introduces cost and risk faster than it creates value. AI amplifying broken process: If DoR not enforced → faster rework loops Token cost / lack of control: No real-time monitoring → cost escalation risk Skills gap: Teams lack prompt engineering capability and understanding of AI usage Governance & guardrails: Risk of inconsistent outputs and lack of explainability Adoption risk: Low engagement from teams, resistance to new ways of working
Lessons Learned	6 Key Lessons: Fix process before AI: AI amplifies existing dysfunction Upstream is the constraint: Poor intake → rework loops → delivery inefficiency Definition of Ready is critical: Must be explicit, measurable, enforced Human-in-the-loop is essential: AI prepares, humans approve Tooling alone is not the solution: Operating model + behaviour change required Training is urgent: Teams need prompt engineering skills and understanding of AI capabilities

Lessons Learned by Service Line

Assurance

Testing is Critical: Testing of AI-native process redesign is critical - uncovers new problems requiring iterative refinement
High-Autonomy Gains: High-autonomy steps deliver immediate efficiency gains
Human-in-the-Loop: Moderate-autonomy stages require explicit human-in-the-loop controls for quality
Token Baseline: Need well-established baseline for monthly token estimate by role
Contract Considerations: Token cost/coverage agreement with suppliers needs consideration in contracts/SOWs
Example - UX Concept: Initial flow had no standard way to share prototype. Updated: PM builds prototype → Figma → XD Lead → Tech Lead → sandbox → business testing

Consulting

Use the Right Tool:
- GitHub Copilot: Best for inline code suggestions
- MS Copilot: Best for general research, summarization
- Factory AI: Reserve for full workflow orchestration only
- Not every task needs Droid - use lighter tools first
Choose the Right Model: Use /model in Droid to select based on task complexity:
- Simple tasks: 0.2-0.4 tokens
- Medium tasks: up to 0.8 tokens
- Complex tasks: 1-2 tokens
Optimize Prompts:
- State your goal clearly in the first prompt
- Mention only the relevant function/class, not the whole file
- Use #file mentions instead of pasting code
- Start a new chat for each new task
- Provide schema or 2-3 row sample instead of full data dumps

Fabric

Early Socialization: Importance of early socialization with delivery teams before introducing AI-enabled model
Remove Blockers First: Remove non-AI process blockers before introducing AI - existing inefficiencies compound with automation
Real Execution Testing: Validate AI-native processes through real execution rather than design alone
Selective AI Usage: Selectively use AI agents to balance efficiency and cost - not every task needs heavy AI
Measure Outcomes: Success measured by value delivered, not volume of features - focus on cycle time, throughput, delivery health

Tax - 7 Key Lessons from Way We Work

1. Tools Alone Don't Change Outcomes - Workflows Do: Adoption stalls when teams are left to "figure it out themselves"
2. Agentic Delivery Requires Explicit Role Design: Agents must be treated as first-class delivery participants with defined scopes
3. Cost and Token Economics Become Material Very Quickly: Always-on heavy agent contexts create exponential cost growth
4. Workflow Coverage is More Effective Than Full SDLC Modeling: Start with high-impact workflows, not comprehensive coverage
5. Upskilling is a Prerequisite, Not a By-Product: Capability building must run in parallel with workflow rollout
6. Measurement Must Shift From Activity to Outcomes: Static KPIs do not fit rapidly evolving agentic systems
7. "Way We Work" Must Be Versioned and Continuously Evolving: Positioned as living initiative that evolves with models, tooling, economics

EY Parthenon - 6 Key Lessons

1. Fix Process Before AI: AI amplifies existing dysfunction - broken processes lead to faster rework loops
2. Upstream is the Constraint: Poor intake → rework loops → delivery inefficiency
3. Definition of Ready is Critical: Must be explicit, measurable, and enforced as a hard gate
4. Human-in-the-Loop is Essential: AI prepares artefacts; humans approve at gates
5. Tooling Alone is Not the Solution: Operating model + behaviour change required for success
6. Training is Urgent: Teams need prompt engineering skills and understanding of AI capabilities

Early Risks by Service Line

⚠ Critical Governance Notice (Consulting)

Effective governance of AI tool usage is now a delivery-critical control. Current signals indicate risk from insufficient guardrails and limited real-time visibility into token consumption, which can lead to unplanned cost escalation and rapid depletion of available investment funding. This creates a direct threat to delivery continuity, including the possibility of pausing usage if budget coverage cannot be confirmed or reducing headcount.

Assurance Risks

Token cost vs. efficiency gain may be cost prohibitive depending on location/quality expectation
Semantic accuracy and overfitting risk in automated processes
Alignment across teams on process, tools, source of truth, and handoffs
Lack of clarity to individuals due to activity overlap/segregation of duties leading to duplicative work

Consulting Risks

Inadequate role-based (including seniority) usage controls driving higher tool costs
Lack of real-time monitoring of token consumption (product/eng code/user/consumed tokens)
Management of token limits at product level including notifications
Defining the right strategy in each phase of PDLC on how to use the tools
Delivery continuity at risk if budget coverage cannot be confirmed
Urgent need for competency-based training and learning plans

Fabric Risks

Adoption challenges if teams are not sufficiently socialized on the AI-enabled model
Dependency on Fabric Portal onboarding for full metrics visibility
Risk of over-using heavy AI agents where lighter tools would be more effective

Tax Risks

Unpredictable tool/token costs - cost governance must be built into workflows from day one
Lack of continuous upskilling of resources - capability building is prerequisite
Continuously addressing cross-cutting concerns: Cost, Security, Performance, Reliability, Quality
Guardrails (what agents can access, execute, transmit) are mandatory
Skills-based agent composition required to reduce token cost and improve quality

EY Parthenon Risks

AI amplifying broken process: If DoR not enforced, AI creates faster rework loops
Token cost / lack of control: No real-time monitoring leads to cost escalation risk
Skills gap: Teams lack prompt engineering capability and understanding of AI usage
Governance & guardrails: Risk of inconsistent outputs and lack of explainability
Adoption risk: Low engagement from teams and resistance to new ways of working

AI Tools by Service Line

Assurance (5 Tools + 3 Integrations)

Tools:

ASU Studio (Factory.ai wrapper) Factory.ai M365 Copilot EYQ GitHub Copilot

Integrations:

Azure DevOps (backlog) GitHub (code) Playwright (test automation)

Consulting (10 Tools)

Figma Motif Experience Studio (in-house) Replit AHA ADO + Skadoosh MCP GitHub Factory.ai (Droid) Cursor GitHub Copilot CLI

Fabric (7 Tools)

Factory.ai DROID CLI GitHub GitHub Actions SL Assist Figma GitHub Copilot

Tax (9 Tools)

Aha! M365 Copilot Figma Figma Make/Replit Azure DevOps GitHub GitHub Copilot Factory.AI Azure Copilot

EY Parthenon (6 Tools)

Factory.ai ADO-integrated Agents GitHub Copilot M365 Copilot / ChatGPT / Claude Figma GitHub

Integrations: Aha!/Intake → ADO, ADO → Factory.ai, Figma ↔ ADO

Agents & Droids Defined

Assurance Agents

Agents defined for each PDLC phase
AI Potential mapping per activity
AI Powered Process details
Agentic approach with human-in-the-loop controls
Playwright integration for test automation

Consulting Droids (8)

Feature/User Story Creator
Code Generator
Code Reviewer
Test Case Generator
Test Script Generator
OSS/Aquasec Finding Fixer
UX Reviewer
Product Document Generator

Future Roles (8)

Business Domain Expert
Prompt Engineer - COE
AI Engineer
UX/UI Designer - Curator COE
Solution Architect - Curator COE
QA Automation Specialist
Project Lead
Knowledge Engineer

Fabric Agents

Feature and Story Generation
Solution Design Assistance
Code Review
Test Case Generation
QA Automation
Human-in-the-loop checkpoints for approvals, quality gates, compliance

Tax Agents (22)

Triage & Router Agent
Product Strategy Agent
Planner Agent
Architect Agent
Developer Agent
Code Review Agent
Test Engineer Agent
CI/CD Agent
Infrastructure Agent
Release Manager Agent
Security Agent
Observability Agent
Incident Commander Agent
Data Engineer Agent
Documentation Agent
Analytics Agent
Compliance Agent
Communication Agent
Dependency Tracker Agent
People & Culture Agent
Knowledge Manager Agent
Experiment Agent

EY Parthenon Agents (6 Core)

Upstream (Highest Priority):
Requirement Intake Agent
PRD/Story Generator Agent
DoR Validation Agent (CRITICAL)
Backlog & Flow:
ADO Health Agent
Planning & Sequencing Agent
Build & Test:
Code Generation Agent
QA Automation Agent
Cross-cutting:
Stakeholder Communication Agent

Future agents: PRD optimisation, NFR validation, Risk prediction, Metrics/insights