The first two modules of this book built the Agent Factory: the paradigm shift from writing software to manufacturing AI employees, the technical architecture of agent systems, the spec-driven development methodology, and the Cowork plugin platform that makes deployment possible without writing code.
Module 7 is where those agents go to work.
This module has a deliberately narrow focus: enterprise business workflows. Not AI in general, not consumer applications, not research prototypes, but the specific functions that run a real organisation: finance, legal, sales, supply chain, product management, human resources, operations, productivity, and innovation. Each chapter deploys AI agents into one of these functions with the same discipline that the rest of this book applies to technical systems: real domain knowledge, production-ready configurations, jurisdiction-aware SKILL.md libraries, and exercises that produce deployable outputs, not proofs of concept.
The central commitment of Module 7 is that domain expertise is the scarce resource, not technical capability. The AI infrastructure is built. The Cowork platform is configured. The MCP connectors exist. What makes an agent in the CFO's office genuinely valuable is not the model; it is the twenty years of financial judgment encoded in its SKILL.md. That knowledge belongs to the professional, not the platform. Module 7 teaches professionals to deploy it.
This module is highly cross-functional. It intentionally bridges the gap between technical engineering and corporate operations.
Here is a breakdown of exactly who should read this module and why:
If Modules 1 through 6 of this Digital FTEs Engineering Book is for the coders, Module 7 is for the builders of the modern enterprise. Anyone involved in the strategy, development, compliance, or management of enterprise-grade AI should consider this required reading.
While this module utilizes Claude Cowork platform for rapid deployment and testing, it is crucial to understand that all workflows, knowledge extraction methodologies, and SKILL.md libraries developed in this module are strictly platform-agnostic. We deliberately architect your domain expertise around open industry standards (such as the emerging protocols documented at https://agentskills.io/home and the Model Context Protocol (MCP)) to ensure your enterprise logic is never locked into a single vendor's ecosystem. By structuring tacit knowledge into standardized, universally readable formats, the agents and skills you build become highly portable intellectual property. They can be seamlessly exported, integrated, and executed across any major agentic framework or enterprise orchestration tool that supports these standards, including Claude Code, Claude Cowork, Gemini CLI, OpenAI Codex, Cursor, and Microsoft Copilot. This design philosophy ensures that the business value you create today remains a durable, adaptable asset regardless of how the underlying AI infrastructure evolves.
Running through all sixteen chapters of Module 7 is a single principle that every student must internalise before deploying agents in any high-consequence domain:
The agent automates execution. The professional makes judgments.
This is not a limitation of current AI capability. It is the correct architecture for enterprise deployment. The IFRS 9 SICR assessment for a borderline credit, whether a loan has suffered a significant increase in credit risk, involves qualitative information no model can fully capture. The SAR filing decision under the Proceeds of Crime Act is a legal obligation that cannot be delegated to software. The closing audit opinion requires a licensed professional's signature. The contract negotiation strategy requires an attorney's judgment.
In every one of these cases, AI agents can dramatically reduce the time and effort required to reach the point where judgment is needed. They can run the ECL calculation, draft the SAR narrative, prepare the audit workpaper, and summarise the contract risk profile. But the professional makes the call.
Every chapter in Module 7 encodes this principle in the SKILL.md files it produces, the escalation protocols it specifies, and the governance frameworks it installs around each domain agent.
Every chapter in Module 7 ends with at least one deployable artefact. Not a prototype. Not a demonstration. A production-ready agent configuration, with a validated SKILL.md library, a defined escalation framework, a Cowork plugin setup, and exercises whose outputs a practitioner can take directly into their organisation.
The artefacts vary by domain. In the Office of the CFO chapters, you will build a complete banking AI plugin stack covering IFRS 9, Basel III/IV, and AML across seven jurisdictions, 23 production-ready SKILL.md files you can deploy immediately. In the Legal & Compliance chapters, you will build a contract lifecycle management agent and Legal Ops workflows that transform the legal function from a bottleneck into a business enabler. In the Growth Engine chapters, you will build sales and marketing agents that encode the judgment of your top performers and scale it across the team.
In every case, the hands-on work is scoped to sessions you can complete in a day. The goal is proof: by the end of each chapter, you will have direct personal evidence that an agent encoding your domain expertise is not a theoretical possibility. It is something you have built, tested against realistic scenarios, and validated against professional standards.
Module 7 spans sixteen chapters across seven sub-modules. The sub-modules follow the value creation architecture of a functioning enterprise: you need foundational methodology before building domain agents (Foundations); financial controls must be right before anything else can be trusted (The Office of the CFO); legal and compliance set the guardrails early (Legal & Compliance); growth depends on sales and marketing at scale (The Growth Engine); your value chain determines whether growth can be delivered (The Product & Value Chain); your people, operations, and productivity sustain the pace (People & Efficiency); and innovation renews the organisation (The Innovation Lab).
The Foundations sub-module establishes the conceptual and technical foundations that every subsequent chapter depends on. It answers three questions practitioners consistently struggle with: What does the enterprise AI landscape actually look like in 2026, and how do I navigate it strategically? How do I architect an agent that can reliably handle the complexity of a real business function? How do I transfer the knowledge locked in expert practitioners' heads into a format that AI agents can execute consistently?
Maps the strategic landscape of enterprise AI in 2026, why adoption stalled in 2024–2025, the platform shift that unlocked deployment, and the Cowork vs. Frontier decision framework. Introduces the four monetisation models that govern how domain agents create financial value and the Organisational AI Maturity Model that tells you which organisations are genuinely ready to deploy.
Key topics: The Pilot Trap and why enterprise AI stalled · Cowork vs. Frontier platform comparison · Knowledge worker centrality · Four monetisation models · Organisational AI Maturity Model · The seven deployment domains
The anatomy of a Cowork plugin, SKILL.md files, the plugin package structure, MCP connectors, the governance layer, the three-way ownership model, and the marketplace. Explains the Persona–Questions–Principles (PQP) Framework, the three-level context hierarchy, and who is responsible for each layer.
Key topics: Cowork plugin anatomy and package structure · The PQP Framework (Persona, Questions, Principles) · Three-level context hierarchy · MCP connector ecosystem · Governance layer and shadow mode · Three-way ownership model · Plugin marketplace
The methodology chapter for the entire curriculum. Describes how to transform the tacit knowledge locked in expert practitioners' heads, and in institutional documents, into the SKILL.md files that make domain agents genuinely useful. Every domain chapter in the subsequent sections applies this method.
Key topics: Method A, expert interview protocol (the five questions that surface tacit knowledge) · Method B, three-pass document extraction from policy manuals · From extraction to SKILL.md translation · Building validation scenario sets · The Validation Loop, shadow mode and graduated autonomy
This sub-module deploys AI agents into financial reporting, compliance, and control: the functions where the cost of an error is not a productivity loss but a restatement, a regulatory breach, a capital adequacy violation, or an enforcement action. Five chapters, each with its own SKILL.md library, Cowork plugin configuration, and jurisdiction-aware exercises.
The sub-module builds systematically. The Foundations chapter establishes the foundational finance agent. Subsequent chapters extend it into the Intent-Driven Financial Architecture. Finally, specialist domains with increasing regulatory complexity are addressed: CA/CPA practice, Islamic finance, and banking-specific AI.
Execution Principle: AI agents in this domain do not replace professional judgment; they automate execution so that professional judgment can be applied where it matters most. The closing audit opinion, the SICR staging decision, and the SAR filing call all require a licensed professional. The agent prepares the ground; the professional decides.
Builds the foundational finance agent covering financial statement analysis, DCF valuation, variance analysis, and FP&A workflows. The only chapter in Module 7 that demonstrates both a Cowork implementation and an OpenAI Frontier implementation side by side, because the finance vertical is where both platforms compete most directly.
Key topics: Financial statement ingestion and normalisation · DCF modelling with scenario analysis · Variance analysis with CFO narrative generation · Budget vs. actuals monitoring with automatic exception flagging · Integration with the finance@knowledge-work-plugins stack
Plugin: finance@knowledge-work-plugins
Extends the foundational finance agent into a methodology for building AI systems that reason about strategic intent, not just data retrieval. The CFO co-pilot: a persistent agent that knows the company's financial history and can connect financial data to business narrative automatically.
Key topics: The IDFA framework, from data query to strategic synthesis · Forward-looking scenario planning agents · Connecting financial data to business narrative automatically · Integration with ERP systems and BI tools via Cowork connectors
Plugin: finance@knowledge-work-plugins + custom connectors
Deploys AI across the full range of CA/CPA practice areas: audit, tax, advisory, and client service. Covers the governance principle that is absolute in this domain: the closing opinion always requires a licensed professional, and builds the SKILL.md library that makes the agent genuinely useful up to that boundary.
Key topics: Audit workpaper automation and analytical procedures at scale · Tax provision computation and disclosure drafting · Client onboarding, engagement letter, and KYC workflows · Continuous accounting, real-time close instead of month-end batch · Practice management, WIP tracking, billing, capacity planning
Plugin: finance@knowledge-work-plugins + audit plugin
The most comprehensive treatment of Islamic finance AI in any curriculum. 16,421 words, 26 SKILL.md files, 7 jurisdiction overlays (Bahrain, Qatar, Malaysia, Saudi Arabia, UAE, UK, Pakistan), and 14 exercises including a full Shariah audit trail and IFRS/FAS 9 interaction analysis. Downloadable Islamic Finance Cowork Skills Library included.
Key topics: Full product library, Murabaha, Ijarah, Diminishing Musharaka, Sukuk, Takaful, Zakat · AAOIFI and IFSB standard compliance and automated Shariah screening · GCC cross-border transaction structuring · Islamic banking vs. conventional banking AI architecture
Plugin: islamic-finance@knowledge-work-plugins (26-file library, 7 jurisdiction overlays)
Full-treatment banking AI across the three regulatory pillars that determine whether a bank is safe, accurate, and clean: IFRS 9 Expected Credit Loss (accounting accuracy), Basel III/IV capital adequacy (solvency), and AML/KYC financial crime prevention. Includes a dedicated section on bank reconciliation, nostro reconciliation, the IFRS 9 four-way provision tie-out, and suspense account control. 14 exercises and a downloadable 23-file Banking Cowork Skills Library across 7 jurisdictions.
Key topics:
Plugin: banking@knowledge-work-plugins (23-file library, 7 jurisdiction overlays: UK, EU, USA, Australia, Singapore, UAE, Pakistan)
This sub-module addresses the function that sets the boundaries within which all enterprise AI operates: legal and compliance. This sub-module is positioned immediately after the financial controls deliberately; legal is an enabling function, not just a constraint. Enterprises that can execute legal processes faster move faster. Enterprises with more effective compliance AI can operate in more jurisdictions with greater confidence.
This sub-module gives significant treatment to Legal Operations Agents: the emerging practice of deploying AI agents specifically within legal department workflows. Legal Ops Agents manage contract lifecycles, route documents for review, flag non-standard clauses, monitor regulatory changes, and maintain the compliance calendar. They transform the legal function from a bottleneck into a business enabler.
Builds AI agents that transform the legal function, from contract lifecycle management through IP protection to regulatory compliance monitoring. The governance principle in this chapter is non-negotiable: certain decisions must always involve a qualified attorney regardless of how accurate an agent becomes.
Key topics: Contract Lifecycle Management (CLM), drafting, redlining, clause library, approval routing · Legal Ops Agents: the emerging practice of AI in legal department workflows · NDA and standard agreement automation · IP protection, patent landscape monitoring, trademark watch, prior art search · Regulatory compliance monitoring across jurisdictions · Risk and compliance tracking, policy gaps, audit readiness, regulatory calendar
Plugin: Legal Ops plugin (built in chapter)
This sub-module addresses the function that determines whether an enterprise grows or stagnates: go-to-market execution. Sales and marketing are domains where AI leverage is exceptionally high, because the core activities—identifying the right prospects, crafting the right message, analysing campaign performance—are pattern-recognition and personalisation tasks at which large language models excel.
The central insight of this sub-module is a reframing: AI does not replace salespeople or marketers. It democratises the capabilities of the best ones. The intuition that a top 1% sales performer has about which prospects are ready, which messages land, and which objections signal genuine interest—that judgment, encoded in a SKILL.md library, can scale across the entire team.
Builds AI agents that scale the judgment of top performers across the entire sales and marketing organisation, from prospecting through pipeline management to campaign optimisation.
Key topics: Prospecting agents, ICP matching, lead scoring, account prioritisation · CRM enrichment, automatic contact data, company research, intent signals · Personalised outreach ghostwriting based on prospect's communication style and history · Campaign planning, content creation, and cross-channel performance analysis · RevOps intelligence, pipeline forecasting, churn prediction, revenue attribution
Plugin: sales@knowledge-work-plugins + marketing@knowledge-work-plugins
This sub-module covers the two domains that connect customer demand to product delivery: supply chain and product management. The fundamental challenge of operational AI is that data lives in multiple systems, decisions must be made in near-real-time, and the cost of error hits the P&L immediately. This sub-module addresses the product manager's core challenge: synthesising customer feedback, technical constraints, business priorities, and market signals into decisions about what to build next.
Deploys AI agents across the end-to-end supply chain, from vendor selection and purchase order management to invoice reconciliation and logistics optimisation.
Key topics: Vendor management agents, supplier risk scoring, performance monitoring, relationship history · Purchase order and invoice reconciliation, three-way match automation at scale · Logistics optimisation, route planning, carrier selection, duty and compliance · Demand forecasting integrating sales pipeline data with inventory planning · Supplier communication, automated RFQ, PO acknowledgement, dispute resolution
Plugin: operations@knowledge-work-plugins
Builds AI agents that transform product management from a reactive coordination function into a proactive strategic capability, from discovery through to stakeholder communication.
Key topics: Feature specification writing, user story generation from discovery notes · Roadmap planning, priority scoring using RICE, WSJF, and custom frameworks · User research synthesis, thematic analysis of interviews, surveys, support tickets · Stakeholder communication, automated sprint updates, release notes, executive summaries · Market and competitive intelligence, continuous monitoring and synthesis
Plugin: product-management@knowledge-work-plugins
This sub-module addresses the infrastructure that determines whether an organisation can execute consistently at scale: its people systems, its operational processes, and the productivity environment in which all knowledge work happens. HR builds the knowledge layer, Operations builds the process layer, and Productivity integrates them into a coherent agentic office.
Builds AI agents that preserve institutional memory, accelerate onboarding, and automate the administrative overhead of human resource management, without losing the human dimension of people operations.
Key topics: Employee onboarding agents, personalised 30-60-90 day plans, document collection, system access · Policy synthesis, converting policy libraries into searchable, conversational knowledge bases · Talent matching, internal mobility, skills gap analysis, succession planning · Internal knowledge base agents, answering HR questions without ticket queues · Performance review automation, template generation, sentiment analysis, calibration support
Plugin: human-resources@knowledge-work-plugins
Builds AI agents that make operational processes visible, documented, optimised, and compliant, transforming operations from a reactive function into a continuous improvement engine.
Key topics: Process documentation agents, turning interviews into SOPs and workflow diagrams · Vendor management, contract tracking, renewal alerts, performance dashboards · Change management, impact assessment, communication planning, rollout monitoring · Compliance tracking, policy adherence, audit preparation, regulatory calendar management · Operational KPI monitoring, exception alerting when metrics breach thresholds
Plugin: operations@knowledge-work-plugins
The integration chapter. Builds the workplace AI layer that knows your organisation's people, projects, terminology, and priorities, and connects all the domain agents into a coherent agentic office environment. This is the chapter where the domain agents stop being point solutions and start working as a system.
Key topics: Task management agents, priority scoring, deadline tracking, meeting follow-up extraction · Workplace memory, persistent context about projects, decisions, and relationships · The agentic daily briefing · Visual dashboard, project status, team capacity, and risk indicators · Cross-domain orchestration: how the domain agents work together
Plugin: productivity@knowledge-work-plugins
The Innovation Lab sub-module is the culmination of Module 7. The student who has completed the previous chapters can build, deploy, and govern domain-specific AI agents across every major enterprise function. This final chapter asks a different question: what happens when that capability is applied not to optimising an existing enterprise, but to creating a new one? The sub-module draws on domain agents from across the entire curriculum and shows how an AI-native startup is structurally different from a traditional one.
Combines Lean Startup, Design Thinking, and Agile with AI-accelerated execution, for both the intrapreneur within a large enterprise and the founder building from scratch.
Key topics: Lean Startup AI agents, hypothesis generation, MVP specification, experiment design · Design Thinking agents, user research synthesis, persona building, journey mapping · Agile AI agents, backlog generation, sprint planning, retrospective synthesis · Business plan writing, AI-assisted financial modelling, market sizing, competitive analysis · Investor pitch preparation, narrative structuring, financial summary, Q&A preparation · Funding strategy, grant identification, investor targeting, pitch iteration
Plugin: Custom innovation plugin (built in chapter)
All 16 chapters are now published. Chapters marked 📝 First Draft are live but have not yet been tested in Cowork; exercises and plugin integrations will be refined in a future pass.
Three threads run through all sixteen chapters and give Module 7 its coherence:
Thread 1. The SKILL.md Accumulation. Every chapter contributes to a growing domain knowledge library. The Islamic Finance plugin library from Chapter 7.7 is immediately deployable by any bank or professional services firm operating in Islamic finance. The Banking plugin library from Chapter 7.8 covers IFRS 9, Basel III/IV, AML, and seven jurisdiction overlays in production-ready form. Students leave Module 7 with intellectual property, not just knowledge.
Thread 2. The Governance Escalation Framework. Every chapter encodes a precise boundary between what the agent executes and what the professional decides. The credit officer makes the SICR staging call. The MLRO makes the SAR filing decision. The attorney decides the negotiation strategy. The auditor signs the closing opinion. Module 7 systematically builds the professional judgment to know where that boundary is in every domain, which is ultimately more valuable than any individual agent configuration.
Thread 3. The Integration Architecture. The seven sub-modules are designed to work together. Financial data feeds the banking compliance agent. Customer intelligence feeds the product roadmap agent. The HR knowledge base feeds the workplace memory layer. Module 7.15: the Agentic Office is the integration chapter that makes domain agents work as a coherent system rather than isolated point solutions.
Module 7 begins with Chapter 7.1: The Enterprise Agentic Landscape.