Enterprise AI
A practical guide to using AI by department — which products to buy, what to plan, CapEx vs OpEx costs, and governance red lines for finance, HR, and logistics.
Part A
The Evaluation Framework
Before evaluating any AI product, follow this six-step framework. It applies to every department — finance, HR, logistics, and beyond. The framework never changes; only the department-specific details do.
Define the function's problem first
Before evaluating AI products, define what decision or work is being automated or assisted. The Logic Centre applied to enterprise: what is the problem, who are the stakeholders, and what outcome defines success?
- —What decision or work is being automated or assisted?
- —Who are the stakeholders and what do they need?
- —What outcome defines success?
What to plan before AI
Before buying AI, plan data readiness, governance, boundaries, and success metrics. The Logic Centre and Boundaries frameworks applied per department.
- —Data readiness — is the data clean, accessible, and governed?
- —Governance — who approves, who is accountable?
- —Boundaries — what must AI not change?
- —Success metrics — how will value be measured?
Which product categories fit
Map the function's needs to product categories — from general-purpose AI to enterprise platforms, RAG, and automation.
- —General Purpose AI — analysis, drafting, research
- —Automation Platforms — workflow, process automation
- —Enterprise AI tier — Bedrock, Azure OpenAI, Vertex AI
- —RAG / knowledge platforms — policy lookup, internal knowledge
Cost model: CapEx vs OpEx
Subscriptions and API calls are OpEx; infrastructure, custom builds, integration, and data preparation are CapEx. The biggest hidden cost is poor thinking that leads to rework.
- —OpEx — subscriptions, API calls, per-user licenses
- —CapEx — infrastructure, custom builds, integration, data prep
- —Hidden cost — poor thinking and rework
Governance & verification
Define who approves AI-assisted work, what AI may not change (Red Lines / Stone Walls), and how outputs are verified before trust.
- —Who approves — accountability before automation
- —Red Lines — what AI must never change
- —Verification — every output checked before trust
Build vs Buy vs Augment
Should you build new AI solutions, buy existing products, or augment existing systems? This decision should be driven by business value, not technological enthusiasm.
- —Build — when no product fits and the capability is strategic
- —Buy — when a mature product meets the need
- —Augment — when existing systems can be enhanced with AI
Part B
The Function Directory
Each function follows the same six-field template: what AI can do, what to plan, which products, cost structure, governance red lines, and a realistic first step.
§ B1
AI for Finance
The finance function breaks into 7 sub-functions, each with concrete AI use cases. This is the most detailed function block on the page.
The 7 finance sub-functions
Each sub-function has concrete AI use cases. Together, they cover the full finance function.
Financial Analysis Team
- ·Trend analysis
- ·Report generation
- ·Investment research
- ·Scenario analysis
Budgeting & Forecasting
- ·Budget creation
- ·Variance analysis
- ·Forecast modeling
- ·Cost-benefit analysis
Controller
- ·General ledger oversight
- ·Compliance monitoring
- ·Financial reporting
- ·Internal controls
Accounts Payable & Receivable
- ·Invoice processing
- ·Payment reminders
- ·Vendor communications
Stakeholder Communication (CFO)
- ·M&A due diligence
- ·Performance metrics (KPIs)
- ·AP/AR reconciliation
- ·Strategic planning
- ·Financial communication drafting
Audit Team
- ·Risk assessment
- ·Audit planning
- ·Issue documentation
- ·Control testing
Tax Compliance & Strategy
- ·Tax filing assistance
- ·Tax code research
- ·Audit preparation
- ·Tax strategy planning
- —Data sensitivity — financial data is regulated; what can go to general-purpose AI vs what needs a secure enterprise tier (Bedrock, Azure OpenAI)?
- —Ledger integrity — AI must never alter the general ledger, pricing, or financial statements without human approval
- —Regulatory compliance — which jurisdictions apply? (IFRS, GAAP, SOX, local tax codes)
- —Audit trail — every AI-assisted output must be traceable to its source data and prompt
General Purpose AI
ChatGPT, Claude, Gemini
Analysis, drafting, research, methodology guides — the bulk of the use cases
Enterprise AI tier
Bedrock, Azure OpenAI, Vertex AI
When financial data sensitivity requires it
RAG / knowledge platform
Vector DB + LLM
Policy lookup, tax code research, compliance checklists
Automation platform
Zapier, Make, n8n
Invoice processing, payment reminders, report generation
OpEx (recurring)
- ·Claude/ChatGPT subscriptions
- ·API calls
- ·Enterprise tier per-user licenses
CapEx (one-time)
- ·RAG platform setup
- ·Data integration
- ·Custom invoice-processing automation
- ·Compliance audit configuration
- ✕AI must never alter the general ledger without human approval
- ✕AI must never set or change pricing
- ✕AI must never finalize financial statements without human review
- ✕AI must never file taxes without human sign-off
- ✕AI must never make investment decisions — it provides analysis, humans decide
Start with the lowest-risk, highest-time-savings use case: methodology and drafting (trend analysis guides, report templates, variance analysis procedures) using a general-purpose AI. Then expand to RAG for policy/tax code lookup. Then automate invoice processing and payment reminders.
§ B2
AI for HR
AI assists HR with screening, onboarding, policy access, training, and analytics — with red lines protecting hiring, termination, and evaluation decisions.
What AI can do
Resume screening
Onboarding automation
Policy Q&A
Training content generation
Performance analytics
- —Bias auditing — AI screening models can inherit historical bias; audit before deployment
- —Data privacy — GDPR, candidate consent, and data retention compliance
- —Candidate communication transparency — disclose when AI is used in the process
General Purpose AI
ChatGPT, Claude, Gemini
Drafting job descriptions, training content, policy summaries
RAG platform
Vector DB + LLM
Policy knowledge base — employees ask questions, get sourced answers
Automation platform
Zapier, Make, Workato
Workflow: resume → screen → schedule → onboard
OpEx (recurring)
- ·Subscriptions
- ·API calls
CapEx (one-time)
- ·RAG setup
- ·HRIS integration
- ✕AI must never make final hiring decisions without human review
- ✕AI must never terminate employment without human review
- ✕AI must never evaluate performance without human context
Start with the lowest-risk use case: drafting job descriptions and training content using a general-purpose AI. Then deploy a RAG platform for policy Q&A. Then automate the resume-to-schedule workflow — with bias auditing at every step.
§ B3
AI for Logistics
AI assists logistics with routing, forecasting, inventory, document processing, and supplier communication — with red lines protecting safety-critical systems and compliance.
What AI can do
Route optimization
Demand forecasting
Inventory management
Document processing (customs, shipping, bills of lading)
Supplier communication
- —Data quality — historical shipment, demand, and inventory data must be clean and complete
- —Real-time integration — ERP, TMS, and WMS connectivity is required for live optimisation
- —Safety-critical system boundaries — define what AI may not touch
Predictive ML
Custom models, Amazon Forecast
Demand forecasting, demand planning, inventory optimisation
Automation platform
Zapier, Make, n8n
Document processing — customs forms, bills of lading, shipping labels
General Purpose AI
ChatGPT, Claude, Gemini
Supplier communication, drafting, exception handling
OpEx (recurring)
- ·API calls
- ·Cloud ML inference
- ·Compute
CapEx (one-time)
- ·Integration with ERP/TMS/WMS
- ·Custom ML models
- ·Data pipeline construction
- ✕AI must not alter safety-critical routing without human override
- ✕AI must not authorize shipments above value thresholds without approval
- ✕AI must not modify customs or compliance documentation without human review
Start with demand forecasting using historical data — the highest-value, lowest-risk use case. Then automate document processing for customs and shipping. Then optimise routing — with safety-critical boundaries defined first.
Part C — Consultancy
Every enterprise AI initiative needs a structured plan before the first product is bought.
We help organisations design that plan — from problem definition to product selection to governance red lines.
§ FAQ
Frequently Asked Questions About Enterprise AI
01What is enterprise AI?+
Enterprise AI is the use of artificial intelligence across business functions — finance, HR, logistics, operations, and more — with governance, cost discipline, and human accountability built in.
02How do I use AI in my business?+
Start by defining the function's problem, planning data and governance, mapping product categories to needs, and establishing red lines before any product is purchased. See our evaluation framework.
03How much does enterprise AI cost?+
Enterprise AI costs include OpEx (subscriptions, API calls, per-user licenses) and CapEx (infrastructure, custom builds, integration, data preparation). The biggest hidden cost is poor thinking that leads to rework.
04What AI products should my company buy?+
It depends on the function. General Purpose AI (ChatGPT, Claude) handles analysis and drafting; enterprise tiers (Bedrock, Azure) handle sensitive data; RAG platforms handle internal knowledge; automation platforms handle workflows. See our function directory.
05How do I use AI in finance?+
AI in finance assists with financial analysis, budgeting, forecasting, accounts payable/receivable, audit, tax compliance, and CFO-level communication — with governance red lines protecting the general ledger, pricing, and financial statements.
06How do I use AI in HR?+
AI in HR assists with resume screening, onboarding automation, policy Q&A, training content generation, and performance analytics — with red lines preventing AI from making final hiring, termination, or evaluation decisions.
07How do I use AI in logistics?+
AI in logistics assists with route optimization, demand forecasting, inventory management, document processing, and supplier communication — with red lines protecting safety-critical routing and customs compliance.
08What is the difference between AI CapEx and OpEx?+
AI CapEx includes infrastructure, custom builds, integration, and data preparation (one-time investments). AI OpEx includes subscriptions, API calls, and per-user licenses (recurring costs). See our CapEx vs OpEx guide.
09How do you govern AI in an enterprise?+
Enterprise AI governance defines who approves AI-assisted work, what AI may not change (red lines), and how outputs are verified before trust. Governance must exist before automation.
10Should we build or buy AI?+
Build when no product fits and the capability is strategic. Buy when a mature product meets the need. Augment when existing systems can be enhanced with AI. See our Build-Buy-Augment framework.
11What is the first step in enterprise AI?+
The first step is defining the function's problem — what decision or work is being automated or assisted — before evaluating any AI product.
12How long does enterprise AI implementation take?+
Basic AI adoption (subscriptions, drafting) takes days. Structured adoption with governance, RAG, and automation takes weeks to months. Full enterprise transformation takes longer.
13What are the risks of enterprise AI?+
The risks of enterprise AI include data leakage, unverified outputs, governance gaps, cost overruns, and automation of poorly designed processes. All are mitigated by planning before AI.
14How do you measure AI ROI?+
AI ROI is measured by time saved × salary (for productivity gains), cost avoided (for automation), and revenue generated (for growth initiatives) — minus the total cost of AI (OpEx + CapEx + rework).
15What AI should we not use?+
AI should not be used for tasks requiring human judgement, accountability, or where errors carry high consequence — and for problems that don't need AI at all. See our normative guide.
16How can AI help with financial analysis?+
AI assists with trend analysis (step-by-step methodology), report generation (procedural templates), investment research (structured approaches), and scenario analysis (frameworks for comparing financial outcomes).
17How can AI help with budgeting and forecasting?+
AI assists with budget creation (comprehensive component checklists), variance analysis (actual vs budgeted), forecast modeling, and cost-benefit analysis.
18Can AI do accounts payable and receivable?+
AI can assist with invoice processing workflows, automated payment reminders, and vendor communication templates — but final approval of payments remains a human responsibility.
19How can AI help auditors?+
AI assists auditors with risk assessment, audit planning, issue documentation, and control testing — but audit conclusions and sign-offs remain human responsibilities.
20Can AI help with tax compliance?+
AI assists with tax filing preparation, tax code research, audit preparation, and tax strategy planning — but tax filings must be reviewed and signed off by a human.
21How can a CFO use AI?+
A CFO can use AI for M&A due diligence checklists, KPI establishment, AP/AR reconciliation guides, strategic planning, and financial communication drafting — with red lines protecting the general ledger and financial statements.
22Can AI replace a financial controller?+
No. AI can assist a controller with ledger oversight, compliance monitoring, and reporting — but accountability for financial integrity remains a human responsibility.
23Is it safe to use AI with financial data?+
It depends. General-purpose AI (ChatGPT, Claude) should not receive regulated financial data. Enterprise tiers (Bedrock, Azure OpenAI) with data residency and privacy guarantees are required for sensitive financial information.