Agent-Native Enterprise Blueprint

Comware: Enterprise AI/ML Consulting

Generated: 2026-02-06 Company: Comware (comware.com.au) Founded: 1993 (30+ years) Location: Australia Scale: Small-to-Medium consulting firm Stage: Established / Transforming -- practicing what they preach Agent Ecosystem: 152 agents, 159 commands, 718 skills, 19 plugins (comware-plugins marketplace, latest versions, verified June 2026)

Table of Contents

1. Executive Summary
2. Business Analysis
3. Agent-Native Vision
4. Capability Pillars
5. Agent Mapping
6. Workflow Analysis
7. Gap Analysis
8. Agent Architecture
9. Implementation Roadmap
10. Risk Assessment
11. Operating Model
12. Cost Estimate
13. Evidence Base (Observed Usage)
14. Appendix: Full Agent Roster
15. Sources & References
16. Summary Statistics

Executive Summary

Comware is a 30-year enterprise AI/ML consulting firm based in Australia, providing end-to-end services from strategy and advisory through custom solution development to deployment. Over the last two months of instrumented work it has become something more specific and more defensible than "a firm planning to adopt AI agents": its core production process -- how it builds and delivers software and AI for clients -- is already agent-native. This is not aspiration. Across 23 of 54 recent projects (42.6%), Comware delivered through an agent-orchestrated pipeline -- spec validation, planning, adversarial audit and release (spectra-sdd), sprint orchestration (project-engine), and content production (cortex) -- with 876 subagent hand-offs and 280 explicit human-in-the-loop checkpoints (Section 12, Evidence Base). Delivery is ~70% of Comware's revenue, and delivery is the part of the firm that is demonstrably, measurably agent-native today.

This reframes the strategic opportunity. Comware does not merely advise clients on agent-native transformation; it delivers inside one -- and can install the same pipeline in client organisations. The most credible thing an AI consulting firm can show an enterprise buyer is not an automated back office (every competitor claims that) but a production line that is faster, spec-correct, auditable by construction, and secured for client data (Sections 9.6, 10.2). That is what enterprises buy, and it is the one claim Comware can back with receipts rather than projections. Section 2.4 develops this into a concrete "Agent-Native Delivery" offering.

The defensible moat follows. The 152 generic agents in the ecosystem are table stakes any competitor can install; the moat is the delivery methodology encoded into the pipeline and refined across real engagements -- proprietary, battle-tested, and uniquely productizable. The "living showcase" is therefore not the whole org chart -- it is the delivery pipeline, and it already exists.

The remaining functions in this blueprint -- Strategy & Advisory, Sales, Client Engagement, Operations, Finance, Marketing, People, Knowledge, Governance -- are reframed from "the plan" to the expansion roadmap: a sequenced extension outward from the proven delivery core. Advisory work (~30% of revenue) is not yet proven agent-native in instrumented tooling -- partly because it happens in conversations and documents the tools don't record, partly genuine gap. It is therefore the highest-value target for the Section 8.1 pilot, not delivery, which is already demonstrated. Across the 14 capability pillars the agent ecosystem shows strong target-state coverage (78%; see Section 3 for how that is scored), with 10 new agents to build across a three-month rollout (Section 6 / Section 7.1 Tier 5), including an agent-security-posture-manager to own the agent platform's security posture (Section 9.6) -- a prerequisite for handling client data, not an afterthought.

Net thesis: lead with the proven agent-native delivery engine (sell it now, productize it -- Section 2.4); expand outward to advisory and operations (prove it via the Section 8.1 pilot, then scale it). The broad vision is unchanged; only its sequencing and headline move to match where the evidence and the revenue actually are. The forward-looking business case still rests on a productivity hypothesis that is not yet proven -- the ROI projections, KPI targets, and the ~85%-of-workflows automation goal are estimates gated on measured pilot results (Sections 8.1, 11.2, 12), not commitments.

Coverage Score: 78%

How coverage is scored. Coverage is a structured qualitative judgment, not a measured metric. For each pillar we enumerate its core functions (the rows in the Section 4 tables), then rate each function's agent support as Direct (purpose-built agent, weight 1.0), Strong (general agent that applies well, 0.75), Partial (adaptable but not designed for it, 0.5), or None (0). The pillar percentage is the support-weighted share of its functions; the 78% headline is the importance-weighted average across the 14 pillars (Critical pillars weighted ~2x Nice-to-have). The percentages are therefore directional planning estimates accurate to roughly ±10 points -- useful for prioritisation, not precise measurements. They should be re-derived once real usage data exists.

1. Business Analysis

1.1 Business Model

Comware operates a professional services model with the following characteristics:

1.2 Revenue Streams (Inferred)

1.3 Client Engagement Lifecycle

1.4 Competitive Landscape

Comware competes in a market that includes:

• Global consultancies (McKinsey, BCG, Deloitte AI practices) [R7] -- broader brand but less specialized
• Boutique AI firms (numerous Australian startups) -- newer, less track record
• Platform vendors (AWS, Azure, Google professional services) -- tied to their platform
• Offshore AI development shops -- lower cost but less strategic advisory

Comware's differentiation: 30+ years of enterprise trust, independence from platform vendors, end-to-end capability (strategy through deployment), and Australian market intimacy.

1.5 Key Business Challenges (Typical for AI Consulting Firms)

1. Talent scarcity -- AI/ML engineers and data scientists are in high demand [R8]
2. Project estimation risk -- AI projects are inherently uncertain in scope
3. Utilization optimization -- balancing bench time with billable work
4. Knowledge retention -- preventing expertise walkout when consultants leave
5. Scalability -- human-dependent delivery model limits growth
6. Rapid technology change -- AI landscape shifts quarterly
7. Client education -- enterprise clients often lack AI readiness
8. Proof of value -- demonstrating ROI on AI investments

2. Agent-Native Vision

2.1 What "Agent-Native" Means for Comware

An agent-native Comware is one where AI agents are embedded into every operational layer -- not as tools bolted on, but as first-class participants in business processes. The vision has three dimensions:

Dimension 1: Internal Operations Every internal business function -- from sales pipeline management to financial forecasting to knowledge curation -- is augmented or automated by specialized agents. Human consultants focus on relationship building, creative problem-solving, and strategic judgment. Agents handle the repetitive, data-intensive, and coordination-heavy work.

Dimension 2: Client Delivery Acceleration The consulting delivery lifecycle is accelerated by agents. The aim is for strategy assessments that took weeks to take days, and for custom ML development cycles, documentation, testing, and deployment to compress through agent assistance. This does not replace consultants -- the working hypothesis is a 3-5x productivity gain on agent-suitable tasks (analysis, drafting, research), which must be validated by the pilot in Section 8.1 before it is treated as an achieved result rather than a target.

Dimension 3: Living Showcase Comware operates as its own case study. Every client engagement can reference "this is how we run our own business" -- demonstrating agent-native transformation with authentic, battle-tested evidence. This is potentially a strong sales differentiator, provided the internal transformation genuinely succeeds; a showcase that over-promises would damage credibility more than help it.

2.2 Target Operating Model

2.3 The Dual Advantage

Note on these figures. The "Traditional" column reflects Comware's current experience; the "Agent-Native" column states target turnaround times, not yet-measured outcomes. They are estimates to be confirmed against the pilot baseline (Section 8.1). Cycle-time gains depend heavily on data availability and the depth of human review each deliverable requires.

2.4 Productizing the Delivery Pipeline

The Evidence Base (Section 12) shows that the part of Comware already operating agent-native is its delivery engine -- the way it builds and ships software/AI for clients. Across instrumented projects, the same pattern recurs: spec validation, planning, adversarial audit, and release (spectra-sdd), sprint orchestration (project-engine), and content production (cortex), with real subagent orchestration and human-in-the-loop checkpoints. That repetition is not just efficiency -- it is a reusable, hardened Agent-Native Delivery Pipeline, and it is Comware's most defensible asset. Today it is treated as internal tooling; it should be treated as intellectual property and a product (the concrete payload of the ip-asset-manager gap, Section 6).

The offering -- "Agent-Native Delivery" -- in three commercial shapes:

Why clients buy it:

• Risk reduction, not just speed -- spec contracts, adversarial audit, and human-review gates mean fewer escaped defects and a defensible audit trail. For regulated buyers this is the headline.
• Proof, not promise -- Comware runs this on its own client work (Section 12); the reference is itself.
• Governed and secured by design -- the agent-security posture (Section 9.6) and governance model (Section 10.2) ship as part of the product, pre-empting the #1 enterprise objection to agent-native delivery.

Why it is defensible. The base agents are table stakes a competitor can install in an afternoon; what a competitor cannot copy is Comware's encoded delivery methodology -- 30 years of judgment plus the agent-orchestration refinement evidenced in Section 12 -- or its track record running it. Productizing also compounds the moat: every engagement feeds engagement-knowledge-extractor -> ip-asset-manager, making the pipeline better and harder to replicate over time.

What it requires (ties to existing gaps): promote ip-asset-manager from "track reusable assets" to "manage the Delivery Pipeline as the flagship product"; treat agent-security-posture-manager (Section 9.6) as a sellable feature, not just internal hygiene; and shift the pricing motion toward value/risk-based pricing for pipeline delivery (pricing-strategist).

Scope note. This is the agent-native capability Comware can prove today. The strategy/advisory and operations pillars (Sections 3-7) remain the expansion roadmap -- sequenced outward from this proven delivery core and validated by the Section 8.1 pilot, whose highest-value target is the as-yet-unproven advisory frontier.

3. Capability Pillars

Based on analysis of Comware's business model, the AI/ML consulting industry requirements, and the existing agent ecosystem, 14 core capability pillars are identified:

Pillar Overview

4. Agent Mapping by Pillar

Target-state mapping. The match ratings below (Direct / Strong / Partial) and the importance/priority labels describe the intended operating model. They are not observed usage -- Section 12 (Evidence Base) shows that actual ANE use to date skews heavily to the engineering/delivery toolchain, with the strategy/advisory agents largely unexercised in instrumented tooling. Read this section as where Comware is heading; read Section 12 as where it is today.

Pillar 1: Strategy & Advisory Services

Importance: Critical Coverage: Mostly Covered -- 90%

This is Comware's bread and butter. The agent ecosystem has purpose-built agents for every phase of AI strategy advisory.

Gaps:

• No agent for industry-specific AI benchmarking (e.g., "where does this mining company sit vs. peers in AI adoption?")
• Could use a regulatory landscape scanner tuned to Australian industry regulations

Pillar 2: AI/ML Solution Delivery

Importance: Critical Coverage: Fully Covered -- 95%

The ecosystem is exceptionally deep for AI/ML engineering, which aligns perfectly with Comware's technical delivery requirements.

Gaps:

• Minimal -- this is the strongest pillar in the ecosystem
• Could benefit from an AI solution architect role that bridges business requirements to technical architecture (partially covered by system-architect + ai-strategy-advisor)

Pillar 3: Sales & Revenue Generation

Importance: Critical Coverage: Mostly Covered -- 85%

Gaps:

• No agent for consulting-specific lead scoring (rating prospects by AI maturity, budget, strategic fit)
• No referral network manager (tracking and nurturing referral relationships, which drive most consulting revenue)

Pillar 4: Client Engagement Management

Importance: Critical Coverage: Partially Covered -- 60%

Gaps:

• CRITICAL: No engagement lifecycle manager -- tracking engagement phases, milestones, billing triggers, scope changes, and the journey from SOW to closeout
• No scope change manager -- consulting engagements constantly evolve; need systematic scope management
• No client portal coordinator -- managing client access to deliverables, status, and communication

Pillar 5: Consulting Operations

Importance: Critical Coverage: Partially Covered -- 55%

Gaps:

• No engagement profitability analyzer -- tracking margin per engagement, comparing estimates to actuals
• No utilization dashboard agent -- real-time consulting metrics (utilization rate, realization rate, leverage ratio)
• No subcontractor management agent -- managing partner/subcontractor relationships, performance, compliance

Pillar 6: Finance & Financial Planning

Importance: Important Coverage: Fully Covered -- 90%

Gaps:

• Minor -- could use consulting-specific revenue recognition modeling (milestone-based billing, T&M true-ups)

Pillar 7: Marketing & Thought Leadership

Importance: Important Coverage: Fully Covered -- 90%

Gaps:

• No case study generator -- converting completed engagements into publishable case studies (anonymized)
• No speaking engagement manager -- tracking CFPs, conference opportunities, speaker preparation

Pillar 8: People & Talent Management

Importance: Important Coverage: Mostly Covered -- 75%

Gaps:

• No consultant career development agent -- managing skill progression, certification tracking, promotion criteria
• No AI talent market intelligence agent -- tracking compensation benchmarks, skill demand, talent availability in Australian market

Pillar 9: Knowledge Management

Importance: Important Coverage: Mostly Covered -- 80%

Gaps:

• No engagement knowledge extractor -- automatically harvesting reusable artifacts, patterns, and lessons from completed client engagements
• No IP asset manager -- tracking reusable code, models, frameworks, and accelerators developed across engagements

Pillar 10: Engineering & Technology

Importance: Important Coverage: Fully Covered -- 95%

Gaps:

• Minimal -- this is excellently covered

Pillar 11: Governance, Risk & Compliance

Importance: Important Coverage: Fully Covered -- 85%

Gaps:

• Australian Privacy Principles (APPs) specific compliance agent
• AI Act (EU) compliance for clients with European operations [R9] (current as of June 2026 -- phasing in through 2026-2028; re-check the applicable deadline per client)
• Agent-platform security posture -- this pillar covers conventional/client security strategy, but the agent ecosystem's own attack surface (prompt injection, least-privilege, secrets, supply chain) and Comware's SOC 2 / ISO 27001 readiness are addressed separately in Section 9.6 and owned by the proposed agent-security-posture-manager

Pillar 12: Innovation & R&D

Importance: Nice-to-have (but strategically valuable) Coverage: Mostly Covered -- 80%

Gaps:

• No accelerator/IP development tracker -- managing the pipeline of reusable accelerators being built across engagements
• No patent/trade-secret monitor -- protecting Comware's proprietary methodologies

Pillar 13: Partnership & Ecosystem

Importance: Nice-to-have Coverage: Mostly Covered -- 75%

Gaps:

• No technology partner program manager -- managing certifications, co-sell motions, MDF programs with AWS/Azure/GCP
• No channel partnership manager -- managing reseller/referral relationships

Pillar 14: Australian Market Compliance

Importance: Important Coverage: Partially Covered -- 45%

Regulatory currency -- current as of June 2026. The legislation referenced in this pillar is actively changing: the Privacy Act 1988 is mid-reform [R1], and the Fair Work Act employee/contractor test changed on 26 August 2024 [R3]. The APRA/ASIC [R4] and AI Ethics Principles [R5] references are stable. Any agent built against these (especially australian-privacy-compliance) must re-verify the law at build time -- see Sources & References for dated specifics.

Gaps:

• CRITICAL: No Australian Privacy Principles (APPs) compliance agent [R1]
• No Australian employment law advisor -- Fair Work Act, contractor vs. employee, visa requirements [R3]
• No APRA/ASIC compliance monitor -- for clients in Australian financial services [R4]
• No Australian AI ethics framework agent -- Australia's AI Ethics Principles alignment [R5]

5. Workflow Analysis

5.1 Daily Operations

Morning Briefing Workflow

Trigger: Daily, 7:30 AM AEST Agents: chief-of-staff, execution-monitor, stakeholder-update-writer Automation Level: Fully automated, consultant-reviewed

Estimated time: 10 minutes (automated) + 5 minutes (human review)

Client Inquiry Handling

Trigger: New inquiry received (email, web form, LinkedIn) Agents: sales-lead, ai-maturity-assessor, consulting-proposal-writer Automation Level: Semi-automated

Estimated time: 30 minutes (automated) + 15 minutes (human review/send)

5.2 Weekly Cadence

Pipeline Review

Trigger: Every Monday, 9:00 AM AEST Agents: sales-lead, forecasting-analyst, competitive-intelligence-analyst Human touchpoints: Partner review of pipeline and forecast

Estimated time: 45 minutes (automated) + 30 minutes (human meeting)

Content & Thought Leadership Planning

Trigger: Every Wednesday Agents: content-strategist, trend-spotter, tech-paper-writer Human touchpoints: Consultant review and authorship

Estimated time: 2 hours (automated) + 1 hour (human refinement)

5.3 Engagement Lifecycle Workflows

New Engagement: Discovery to Proposal

Trigger: Qualified lead moves to discovery Agents: discovery-protocol, ai-maturity-assessor, ai-use-case-analyst, consulting-proposal-writer, pricing-strategist Human touchpoints: Client meetings, proposal sign-off

Estimated time: 3-5 days (traditional: 2-3 weeks)

AI Strategy Engagement Delivery

Trigger: Engagement kicks off after SOW signing Agents: ai-strategy-advisor, ai-maturity-assessor, ai-workshop-facilitator, ai-use-case-analyst, ai-ethics-auditor Human touchpoints: All client-facing interactions, final recommendations

Estimated time: 4 weeks (traditional: 8-12 weeks)

5.4 Monthly Cadence

Financial Review

Trigger: First Monday of each month Agents: financial-modeler, budgeting-specialist, revenue-model-analyst, forecasting-analyst

Estimated time: 2-3 hours (automated) + 30 minutes (human review)

Practice Health Check

Trigger: Monthly Agents: consulting-staffing-specialist, workforce-planner, process-optimizer

5.5 Quarterly Cadence

Strategic Review & Planning

Trigger: Quarterly Agents: strategy-lead, corporate-strategy-analyst, strategic-planning-facilitator, okr-designer

5.6 Event-Driven Workflows

New Technology Response Workflow

Trigger: Major AI technology announcement (new model release, framework, regulation) Agents: technology-radar-monitor, trend-spotter, research-synthesizer, content-strategist

Estimated time: 4-6 hours (traditional: 1-2 weeks to respond to market shifts)

Client Escalation Workflow

Trigger: Client satisfaction issue or delivery risk identified Agents: risk-analyzer, crisis-communications-manager, human-handoff-manager

6. Gap Analysis

6.1 Coverage Summary

6.2 Critical Gaps (Must Build)

6.3 Important Gaps (Should Build)

6.4 Nice-to-Have Gaps (Could Build Later)

6.5 Gap Agent Specifications

engagement-lifecycle-manager

• Purpose: Tracks consulting engagements from signed SOW through delivery, billing, closeout, and into ongoing relationship management. Manages milestones, deliverables, billing events, team assignments, scope changes, and client satisfaction through the full engagement lifecycle.
• Pillar: Client Engagement Management
• Tools needed: Read, Write, Edit, Glob, Grep, Bash, WebSearch
• Inputs: SOW/contract details, project plans, timesheet data, billing records, client feedback
• Outputs: Engagement status dashboards, milestone tracking, billing triggers, scope change logs, engagement health scores, closeout checklists
• Connects to: consulting-staffing-specialist, delivery-readiness-assessor, stakeholder-update-writer, financial-modeler, customer-success-lead
• Build effort: Medium -- requires integration with project data sources and billing systems; the core logic is workflow orchestration

engagement-profitability-analyzer

• Purpose: Analyzes financial performance of consulting engagements. Compares estimated vs. actual hours, tracks margin by engagement, identifies profitability patterns, and provides early warnings on engagements trending below target margin.
• Pillar: Consulting Operations
• Tools needed: Read, Write, Edit, Glob, Grep
• Inputs: Engagement budgets, timesheet data, billing records, consultant cost rates
• Outputs: Engagement P&L statements, margin trend analysis, profitability alerts, rate realization reports
• Connects to: financial-modeler, consulting-staffing-specialist, budgeting-specialist
• Build effort: Small -- primarily analytical logic over financial data

practice-metrics-dashboard

• Purpose: Generates real-time consulting practice metrics including utilization rate, realization rate, pipeline-to-revenue conversion, leverage ratio, and bench cost. Provides the KPIs that consulting leaders need for operational decision-making.
• Pillar: Consulting Operations
• Tools needed: Read, Write, Edit, Glob, Grep
• Inputs: Timesheet data, billing data, pipeline data, headcount data
• Outputs: Practice dashboard reports, utilization heat maps, trend analysis, alerts on metric deviations
• Connects to: consulting-staffing-specialist, financial-modeler, metrics-dashboard-creator
• Build effort: Small -- combines data from existing sources into consulting-specific visualizations

australian-privacy-compliance

• Purpose: Ensures compliance with Australian Privacy Principles (APPs) under the Privacy Act 1988 [R1]. Assesses data handling practices, provides guidance on Notifiable Data Breaches (NDB) scheme [R2], and helps clients navigate Australian data sovereignty requirements.
• Pillar: Australian Market Compliance
• Tools needed: Read, Write, Edit, Glob, Grep, WebSearch, WebFetch
• Inputs: Data handling policies, system architectures, data flow diagrams
• Outputs: APP compliance assessments, NDB readiness reports, privacy impact assessments, data sovereignty recommendations
• Connects to: data-privacy-engineer, compliance-checker, ai-ethics-auditor
• Build effort: Medium -- requires deep knowledge of Australian privacy legislation and recent amendments. Currency: as of June 2026 the Privacy Act 1988 is under active reform (2024 amendments + further 2025-26 tranches); this agent must be built against the law in force at build time, not this date. [R1] [R2]

engagement-knowledge-extractor

• Purpose: Automatically harvests reusable knowledge from completed consulting engagements. Extracts methodologies, solution patterns, technical approaches, lessons learned, and reusable artifacts. Feeds the organizational knowledge base to prevent knowledge walkout.
• Pillar: Knowledge Management
• Tools needed: Read, Write, Edit, Glob, Grep
• Inputs: Engagement deliverables, project documentation, retrospective outputs, code repositories
• Outputs: Knowledge base entries, reusable artifact catalog, methodology improvements, pattern library updates
• Connects to: knowledge-curator, knowledge-base-architect, lessons-learned-extractor
• Build effort: Medium -- requires intelligent extraction and classification of diverse engagement artifacts

ip-asset-manager

• Purpose: Tracks Comware's intellectual property including reusable code libraries, ML model templates, data pipeline accelerators, assessment frameworks, and proprietary methodologies. Manages asset lifecycle from creation through utilization tracking.
• Pillar: Knowledge Management / Innovation
• Tools needed: Read, Write, Edit, Glob, Grep, Bash
• Inputs: Code repositories, framework documents, asset metadata
• Outputs: IP asset registry, usage tracking reports, asset health assessments, licensing recommendations
• Connects to: knowledge-curator, open-source-auditor, license-compliance-checker
• Build effort: Small -- primarily registry and tracking logic

consulting-lead-scorer

• Purpose: Scores and qualifies incoming consulting leads based on AI/ML consulting-specific criteria: client AI maturity (high maturity = lower sales effort), budget authority signals, strategic fit with Comware's strengths, competitive dynamics, and expansion potential.
• Pillar: Sales & Revenue
• Tools needed: Read, Write, Edit, Glob, Grep, WebSearch, WebFetch
• Inputs: Lead information, company research, industry signals
• Outputs: Lead scores with rationale, qualification recommendations, suggested engagement approach
• Connects to: sales-lead, ai-maturity-assessor, crm-optimizer
• Build effort: Small -- scoring logic over lead data with web research enrichment

case-study-generator

• Purpose: Converts completed engagement data into polished, anonymized case studies for marketing use. Extracts the challenge, approach, solution, and outcomes into compelling narratives that demonstrate Comware's capabilities.
• Pillar: Marketing & Thought Leadership
• Tools needed: Read, Write, Edit, Glob, Grep
• Inputs: Engagement summaries, deliverable documents, outcome metrics, client approvals
• Outputs: Case study documents (long-form and summary), social media excerpts, website content
• Connects to: content-strategist, executive-summary-writer, brand-consistency-checker
• Build effort: Small -- content generation with structured inputs

scope-change-manager

• Purpose: Manages the scope change process for consulting engagements. Tracks change requests, assesses impact on timeline and budget, generates change orders, and maintains an audit trail of scope evolution.
• Pillar: Client Engagement Management
• Tools needed: Read, Write, Edit, Glob, Grep
• Inputs: Original SOW, change requests, impact assessments
• Outputs: Change order documents, impact analysis, scope evolution timeline, budget impact reports
• Connects to: engagement-lifecycle-manager, consulting-proposal-writer, risk-analyzer
• Build effort: Small -- structured workflow over document data

agent-security-posture-manager

• Purpose: Owns the security posture of the agent-native operation itself (see Section 9.6). Maintains the agent threat model, enforces least-privilege and tool/data allow-lists per agent, monitors for prompt-injection and data-egress events, tracks agent/skill supply-chain provenance, and manages Comware's own SOC 2 / ISO 27001 readiness evidence. Distinct from security-lead (client-facing security strategy) -- this agent governs the internal agent platform.
• Pillar: Governance, Risk & Compliance (agent-platform security)
• Tools needed: Read, Write, Edit, Glob, Grep, Bash, WebSearch
• Inputs: Agent/skill registry and provenance, agent permission/tool configs, access logs, secret-management policy, data-classification policy, incident reports
• Outputs: Agent threat model, least-privilege matrix, supply-chain/provenance audit, prompt-injection & egress monitoring rules, SOC 2 / ISO 27001 readiness checklist, security-posture dashboard
• Connects to: security-lead, zero-trust-architect, penetration-tester, data-privacy-engineer, guardrail-engineer, audit-log-architect
• Build effort: Medium -- combines policy, monitoring config, and registry auditing; the controls it enforces (isolation, least-privilege, secrets) are foundational and should precede giving agents access to real client data

7. Agent Architecture

7.1 Complete Agent Roster

Tiers are target-state priorities, not observed usage. The "Essential/Important" labels reflect intended importance to the consulting model. Observed transcripts (Section 12) show today's actual load-bearing usage is the engineering/delivery toolchain (spectra-sdd, project-engine) rather than the Tier-1 strategy/sales agents -- partly aspiration, partly corpus skew. Sequence activation by both intended priority and demonstrated usage.

Tier 1: Core Consulting Agents (22 agents)

These agents directly support Comware's core consulting business and should be activated first.

Tier 2: Supporting Business Agents (35 agents)

These agents support revenue generation, operations, and growth.

Tier 3: Orchestration & Coordination Agents (10 agents)

These agents coordinate workflows and manage the agent ecosystem.

Tier 4: AI/ML Technical Deep-Dive Agents (25 agents)

Available for client delivery when needed.

Tier 5: New Agents to Build (10 agents)

7.2 Agent Interaction Architecture

8. Implementation Roadmap

8.1 Baseline & Validation Pilot (Week 0 -- prerequisite)

Before scaling, capture a current-state baseline and validate the core productivity hypothesis on a small, controlled sample. Every target in Section 10.3 and every figure in Section 11 should be re-anchored to these measured numbers. The pilot targets the advisory frontier specifically -- delivery is already evidenced (Section 12), advisory is not. A pre-registered measurement plan (metrics, counterfactual design, and fixed Go/No-Go thresholds) is detailed in docs/whitepaper-review/PILOT-MEASUREMENT-PLAN.md.

Current-state baseline to capture (from existing systems):

Pilot design: Run agent augmentation on 2-3 live engagements and one sales pursuit for 4-6 weeks, with a matched set of non-augmented work as comparison. Measure consultant hours per deliverable, cycle time, rework rate, and a quality score from senior-consultant review, pre vs. post.

Go / no-go criteria to proceed to full rollout:

• Measured productivity gain on agent-suitable tasks >= 15% (the ROI base case assumes ~30%; below 15% the economics require revisiting)
• No quality regression in senior-consultant review
• No client-data or confidentiality incident
• Consultant-reported willingness to continue

If criteria are not met, pause and remediate (prompt quality, agent selection, workflow design) before scaling. The roadmap below assumes the pilot passes.

8.2 Data & System Integration (the practical critical path)

The agent definitions are the visible work; connecting agents to Comware's data is the harder, rate-limiting work and is frequently underestimated. The operations, finance, and client-engagement agents (practice-metrics-dashboard, engagement-profitability-analyzer, engagement-lifecycle-manager) are only as good as their access to timesheet, billing, CRM, and project data. Before those agents deliver value:

This integration layer should be scoped explicitly in Phase 1 and may extend the realistic timeline for the operations/finance agents beyond their "Small/Medium" build labels, which reflect agent logic effort only, not data plumbing. Where live integration is not yet feasible, agents can begin with manual/exported inputs and be wired to live sources incrementally.

Phase 1: Foundation (Weeks 1-2)

Objective: Get core consulting delivery and operations running with agent support.

Activate:

• chief-of-staff -- central coordination
• swarm-orchestrator -- multi-agent workflows
• ai-strategy-advisor -- core delivery agent
• ai-maturity-assessor -- core delivery agent
• ai-use-case-analyst -- core delivery agent
• consulting-staffing-specialist -- resource management
• discovery-protocol -- engagement discovery
• delivery-readiness-assessor -- quality assurance
• knowledge-base-architect -- begin knowledge capture
• knowledge-curator -- begin knowledge organization
• context-manager -- agent context sharing
• human-handoff-manager -- human escalation rules

Build:

• engagement-lifecycle-manager -- critical operational gap
• practice-metrics-dashboard -- operational visibility
• agent-security-posture-manager -- establish least-privilege, data-isolation, and secrets controls (Section 9.6) before agents are given access to real client data

Workflows to Implement:

• Daily morning briefing
• Client inquiry handling
• Engagement discovery workflow

Milestone: Core consulting engagements can be delivered with agent augmentation. Consultants report measurable time savings on assessment and analysis work.

Phase 2: Revenue Engine (Weeks 3-4)

Objective: Accelerate the sales pipeline and proposal process and make pipeline metrics real-time.

Activate:

• sales-lead -- pipeline management
• consulting-proposal-writer -- proposal creation
• pricing-strategist -- engagement pricing
• rfp-manager -- competitive bidding
• competitive-intelligence-analyst -- market monitoring
• competitive-battlecard-creator -- sales enablement
• sales-enablement-specialist -- sales training content
• win-loss-analyst -- deal analysis
• financial-modeler -- revenue modeling
• forecasting-analyst -- financial forecasting

Build:

• engagement-profitability-analyzer -- margin tracking
• consulting-lead-scorer -- lead qualification

Workflows to Implement:

• Weekly pipeline review
• New engagement: discovery to proposal
• Win/loss analysis after deal outcomes

Milestone: Proposal turnaround time reduced by 60%. Pipeline visibility is real-time. Lead qualification is systematized.

Phase 3: Operational Excellence (Month 2)

Objective: Full operational visibility, knowledge management, and financial control.

Activate:

• budgeting-specialist -- budget management
• revenue-model-analyst -- revenue analysis
• content-strategist -- thought leadership pipeline
• tech-paper-writer -- technical content
• seo-specialist -- search optimization
• campaign-executor -- marketing campaigns
• marketing-lead -- marketing strategy
• trend-spotter -- trend identification
• technology-radar-monitor -- technology scouting
• research-synthesizer -- research consolidation
• people-lead -- people strategy
• workforce-planner -- capacity planning
• governance-lead -- risk coordination
• compliance-checker -- regulatory compliance
• data-privacy-engineer -- privacy engineering
• process-optimizer -- operational efficiency

Build:

• engagement-knowledge-extractor -- knowledge harvesting
• ip-asset-manager -- IP tracking
• australian-privacy-compliance -- local compliance
• case-study-generator -- marketing content from engagements

Workflows to Implement:

• Monthly financial review
• Monthly practice health check
• Weekly content and thought leadership planning
• Quarterly strategic review
• New technology response workflow

Milestone: Complete operational visibility. Thought leadership pipeline produces consistent output. Knowledge is systematically captured from every engagement.

Phase 4: Strategic Differentiation (Month 3+)

Objective: Innovation, market leadership, and the "living showcase" for clients.

Activate:

• All remaining Tier 2 and Tier 4 agents as needed for specific engagements
• disruption-strategist -- strategic foresight
• business-model-validator -- new offering validation
• idea-brainstormer, idea-lead -- innovation pipeline
• rapid-validator -- fast concept testing
• partnership-evaluator -- strategic partnership assessment
• scenario-planner -- long-term planning
• Full LLM engineering agent suite for advanced client work

Build:

• scope-change-manager -- engagement scope governance
• Remaining nice-to-have gap agents based on business needs

Workflows to Implement:

• Innovation pipeline (idea to prototype)
• Partnership evaluation process
• Client escalation workflow
• Annual strategic planning process

Milestone (outcome-gated, not date-gated): Month 3 marks the start of Phase 4, not a finish line. Entry to Phase 4 is gated on Phases 1-3 having met their milestones -- in particular a passed validation pilot (Section 8.1), live data integration for the operations agents (Section 8.2), and demonstrated consultant adoption. Phase 4 then runs until Comware reaches its target operating model: the 11 core workflows are agent-augmented with reliable human-review gates, and the operating model itself is mature enough to demonstrate to clients ("Let us show you how we run our own business with AI agents"). Realistically this maturity is a 6-12 month journey beyond month 3; the three-month figure covers reaching a working agent-native baseline, not full maturity.

Implementation Timeline

9. Risk Assessment

9.1 Transformation Risks

9.2 Business Continuity Risks

9.3 Risk Mitigation Framework

9.4 Why This Might Not Work (Devil's Advocate)

A balanced plan must state the case against itself:

• The productivity gains may not materialise at the assumed level. If agent-suitable work is a smaller share of consultant time than assumed, or review overhead eats the savings, the 30% hypothesis -- and the ROI built on it -- weakens substantially. This is why Section 8.1 gates rollout on measured pilot results.
• Freed time may not convert to value. Productivity gains only become ROI if freed hours convert to billable work, won business, or avoided hiring. In a soft demand environment, they may simply become idle bench -- a cost, not a benefit.
• Maintenance burden of a large ecosystem. 152 agents, 159 commands, and 718 skills carry real upkeep cost (prompt drift, model changes, quality variance). A leaner set Comware actually uses may beat a large set it cannot maintain.
• Adoption and culture risk is the most common failure mode. Tooling rarely fails on capability; it fails on adoption. If consultants route around the agents, none of the benefits land.
• Data integration may dominate the timeline (Section 8.2), pushing real value later than the roadmap implies.

9.5 Alternatives Considered

This blueprint recommends a custom agent-native build, but it is not the only option:

The recommendation stands, but it is a genuine choice with a credible do-less alternative -- not a foregone conclusion.

9.6 Agent Security & Threat Model

An agent-native operation introduces an attack surface that conventional IT security does not fully address: agents read untrusted content, act on client data, and call tools on their own. For a firm whose pitch is "we run our business on 150+ agents touching client data," this is the area an enterprise client will scrutinise hardest -- so it is treated explicitly here rather than left implicit in Pillars 10-11. These threats are governed in addition to (not instead of) the conventional controls in Pillar 11 and the data-security row in Section 10.2.

Comware's own security posture is a prerequisite, not an afterthought. The "living showcase" only works if Comware can demonstrate it is itself trustworthy with client data. Enterprise buyers will ask for evidence -- typically SOC 2 Type II and/or ISO/IEC 27001 readiness, a documented agent-security policy, and the data-isolation and least-privilege controls above. These should be established before agents are given access to real client data, not retrofitted after an incident. This gap is owned by the proposed agent-security-posture-manager (Section 6, Important gap #10) and reinforced by security-lead, zero-trust-architect, and penetration-tester for periodic assessment.

10. Operating Model

10.1 Human-in-the-Loop Points

These decisions always require human judgment and cannot be delegated to agents:

10.2 Governance Model

10.3 Metrics & KPIs

Targets below are aspirational until anchored to the current-state baseline captured in Section 8.1. The Baseline column is to be filled from existing systems before rollout so each target is expressed as a delta from today, not an absolute pulled from the air.

11. Cost Estimate

11.1 Monthly Operating Costs

Usage-tier note (revised against observed data, see Section 12). The original $2-5k/month figure assumed lean, on-demand agent invocation. Observed ANE usage in heavy software-delivery sessions (frontier model + large prompt-cache reads) implies an API-equivalent run rate far above that band -- on the order of tens of thousands per month if billed per-token. Actual cash cost depends heavily on the billing model (flat-rate subscription vs. metered API) and on how disciplined invocation is. Plan for two tiers: advisory/light-automation work fits the original band; intensive engineering/delivery work does not. finops-specialist + cost-anomaly-detector should enforce per-engagement budgets, and self-hosted models should be evaluated for the heaviest workloads.

11.2 ROI Projection

ROI scenarios (benefit ÷ cost):

Methodology and caveats. Every benefit line above is a modelled estimate, not measured data. The largest single line -- consultant productivity gain -- assumes the 30% hypothesis (Section 8.1) holds and converts to billable or business-development time; if the pilot shows 15% instead, the productivity benefit roughly halves and the base case falls to ~4x. The estimates assume (a) the firm can convert freed consultant time into revenue or saved hiring, not idle bench; (b) AI API costs stay within the Section 11.1 band; and (c) gap-agent development lands at the lower end of effort. These numbers should be replaced with pilot-derived figures before being used in any external or board commitment. The directional case is nonetheless strong: Comware's primary cost is human time, and agent augmentation leverages that time more effectively even under conservative assumptions.

12. Evidence Base (Observed Usage)

Unlike the projections elsewhere in this blueprint, this section is observed data -- mined from Comware's own Claude Code transcripts (2,050 files → 664 sessions across 54 projects, 2026-03-30 to 2026-06-04). It was produced locally and reported aggregate-only, in line with the Section 9.6 data-handling rules. Full method and figures: docs/whitepaper-review/evidence/EVIDENCE-BASE.md.

What the data confirms:

• Adoption is real: ANE capabilities were used in 23 of 54 projects (42.6%), across 56 sessions.
• The agent-native operating pattern is observable, not theoretical: 876 subagent spawns, heavy automation (Bash/Edit/Write), and 280 human-in-the-loop AskUserQuestion checkpoints -- direct evidence for the Section 10.2 governance model.

What the data reveals -- observed usage diverges from the target mapping: The capabilities actually used are dominated by the software-delivery toolchain -- spectra-sdd (566 invocations, spec-driven development), project-engine (158, sprints), cortex (85, content utilities), git-workflow (21). The Strategy/Advisory, Sales, and AI/ML-delivery agents that Section 7.1 rates "Tier-1 Essential" barely register in this corpus. Two factors explain this, and both are stated honestly: (1) the relevance/priority ratings in Sections 4, 7.1, and the Appendix are target-state, not observed -- they describe intended use; (2) corpus skew -- these are Claude Code transcripts centred on software/spec work, so advisory delivery (workshops, strategy decks, client meetings) that happens outside the tool is under-represented. Net: the engineering/delivery pillars are the proven, load-bearing ANE use today; the strategy/advisory pillars remain unproven within instrumentable tooling and are the priority to validate in the Section 8.1 pilot.

What the data cannot show (and why): the productivity multiplier (the 30% / 3-5x in Sections 2.3 and 11.2) is not derivable from transcripts -- they capture agent-assisted effort but no "without ANE" counterfactual. This remains a hypothesis for the Section 8.1 forward pilot. Observed session "duration" is also unreliable (idle time inflates it) and is deliberately not used as a cycle-time metric.

What the data corrects: observed token intensity (frontier model + ~38 B prompt-cache-read tokens over the window) materially exceeds the original Section 11.1 cost assumption -- hence the usage-tiered revision there.

How to read this blueprint in light of the evidence: treat the pillar coverage and relevance ratings as a target operating model, and this section as the observed starting point. The gap between them is the actual transformation work -- and the Section 8.1 pilot is what closes it with measured data.

Appendix: Full Agent Roster

This appendix is a reference index of the full ecosystem by relevance. It intentionally overlaps with the prioritised tier tables in Section 7.1; use Section 7.1 for the activation plan and this appendix for lookup.

All Agents by Relevance to Comware

Legend: (relevance ratings are target-state judgments, not observed usage -- compare with Section 12)

• H = High relevance (directly supports Comware's core business)
• M = Medium relevance (supports operational functions)
• L = Low relevance (available but not primary use case)

AI Consulting & Strategy (H)

Sales & Business Development (H)

Consulting Operations (H)

Client Relationship (H)

AI/ML Engineering (H)

LLM Engineering Suite (H for client delivery)

Finance & Strategy (M-H)

Marketing & Communications (M-H)

Governance, Risk & Compliance (M)

Knowledge & Innovation (M-H)

People & Culture (M)

Orchestration & Meta (H)

Cloud & Infrastructure (M -- for client delivery)

Relevant Commands

Verified June 2026. The commands below were checked against the installed comware-plugins command inventory (see docs/whitepaper-review/evidence/verify_agent_names.py). An earlier version of this table listed intuitive-but-nonexistent commands (e.g. /cos:plan, /llm:eval, /business:evaluate-idea, /research:smart); those were generated names that do not map to real commands and have been replaced with verified equivalents, ordered delivery-first to match the operating reality (Section 12).

Sources & References

External factual and regulatory claims in this blueprint are cited below (markers [R#] appear inline at each claim). Verified June 2026. Note on scope: internal projections -- productivity multipliers, ROI scenarios, coverage percentages, margins, time savings, and cost estimates -- are modelled assumptions, not externally citable facts; they are labelled as estimates in-text and are to be validated against the Section 8.1 pilot rather than cited. The ecosystem counts (152 agents, 159 commands, 718 skills, 19 plugins) and firm facts (founded 1993) are internal inventory figures, verifiable from Comware's own plugin/agent registry rather than from external sources.

Registry verification (June 2026) -- figures corrected. The headline ecosystem counts were verified against the live comware-plugins registry (latest plugin versions, node_modules excluded) and updated to the measured values: 152 agents, 159 commands, 718 skills, 19 plugins. The original 2026-02-06 generation figures (539 agents, 174 commands, 27 skills, 113 plugins) did not reconcile and have been replaced throughout this document, including the derived relevance counts in Summary Statistics (now 78 H / 55 M / 133 mapped, matching the Appendix roster). Counting method: SKILL.md / agents-dir .md / commands-dir .md under each plugin's latest version in the comware-plugins marketplace; "plugins" = distinct plugin packages in that marketplace (58 plugin entries are installed across all marketplaces). Re-run this verification before any external or board use, as the registry changes over time.

Summary Statistics

This blueprint was generated by the Agent-Native Architect using the Comware agent ecosystem (152 agents, 159 commands, 718 skills, 19 plugins; comware-plugins marketplace, verified June 2026). It represents a comprehensive analysis of how an established AI/ML consulting firm can transform into an agent-native enterprise -- becoming both practitioner and exemplar of the technology it delivers to clients.

The dual advantage is clear: Comware does not merely advise on AI transformation. It lives it.