The Practical Core Framework Set for End-to-End AI Solution Engineering

Most AI engagements fail not because the organisation lacks frameworks, but because it has too many of the wrong kind. Teams accumulate strategy canvases, maturity models, scoring formulas and governance checklists until the methodology itself becomes the delivery risk. The practical response is not a larger catalogue. It is a core set: enough structure to run end-to-end AI solution engineering, and little enough that practitioners can actually use it.

This article removes frameworks that are repetitive, overly specialised, primarily academic, or useful only in narrow industries. What remains is a runnable methodology: a spine that governs every engagement, stage frameworks that answer specific decisions, and a final core library of approximately fifty tools that cover the journey without drowning it.

The principle throughout is simple:

Choose one primary tool for each decision. Keep a reference library for edge cases. Never confuse framework coverage with delivery progress.

---

Core methodology spine​

These seven constructs sit above the entire framework library. They are not optional workshop canvases. They are the operating system for the engagement.

#	Spine	Role
1	8D AI Solution Engineering Framework	Structures diagnosis, design and delivery discipline across the problem space
2	VALUE Quality Gate	Forces evidence before investment escalates
3	AI System Lifecycle	Owns the full path from problem to retirement for AI systems
4	Agent Development Lifecycle	Specialises the lifecycle for agents: build, test, deploy, monitor
5	Responsible AI Lifecycle	Makes trust, impact and assurance continuous rather than ceremonial
6	Secure AI Development Lifecycle	Embeds threat modelling, controls and incident readiness into delivery
7	Benefits Realisation Lifecycle	Prevents “model live” being mistaken for business value

Everything else in this article supports a stage of that journey. If a framework cannot map to a spine decision or a stage gate, it does not belong in the core set.

---

1. Mobilise and govern​

Use these frameworks to establish scope, accountability and decision-making before discovery workshops begin. Unclear sponsorship destroys more AI programmes than weak models.

Essential frameworks​

• Project Charter
• RACI
• RAPID
• RAID Log
• Decision Log
• Stakeholder Influence–Interest Matrix
• MECE Issue Tree
• Objectives and Key Results
• Definition of Ready
• Definition of Done

Why these are sufficient​

Together they answer the questions that unblock every later stage:

• What are we trying to achieve?
• What is in and out of scope?
• Who owns the result?
• Who makes each decision?
• What risks and assumptions exist?
• What evidence is required before work progresses?

Avoid unnecessary duplication​

Choose deliberately:

• RACI for work ownership
• RAPID for important decisions
• RAID for risks, assumptions, issues and dependencies

You normally do not need RACI, RASCI, DACI, DARE and ARCI on the same engagement. One ownership model, one decision model, one risk register. Publish them in the first week and revisit them at every stage gate.

Artefacts to leave mobilisation with: signed charter, RACI, RAPID for high-stakes decisions, initial RAID, stakeholder map, and Definition of Ready / Done for the first discovery and design increments.

---

2. Define strategy and strategic alignment​

Use these frameworks to decide why AI should be used and where investment should be directed. Strategy work that starts with a model catalogue is already off-track.

Essential frameworks​

• Strategy Choice Cascade
• Playing to Win
• Three Horizons
• Value Chain Analysis
• Business Model Canvas
• Operating Model Canvas
• Capability-Based Planning
• Wardley Mapping
• Value-Driver Tree
• Scenario Planning
• SWOT
• PESTLE
• Porter’s Five Forces

Most important combination​

For most AI strategy engagements, the strongest combination is:

1. Strategy Choice Cascade
2. Value Chain Analysis
3. Capability-Based Planning
4. Operating Model Canvas
5. Value-Driver Tree
6. Three Horizons

This combination identifies:

• Strategic ambition
• Business areas suitable for AI
• Capabilities that must be developed
• Required organisational changes
• Expected sources of value
• Short-, medium- and long-term initiatives

Use selectively​

SWOT, PESTLE and Porter’s Five Forces are useful for external and competitive analysis, but they should not become the main AI methodology. They inform the cascade; they do not replace it.

Decision to force: where will AI create durable advantage in this value chain, and which capabilities must exist before scale investment is rational?

---

3. Discover users, processes and problems​

Use these frameworks to replace assumptions with evidence. AI programmes that skip discovery usually build elegant solutions to the wrong problem.

Essential frameworks​

• Design Thinking
• Double Diamond
• Jobs to Be Done
• Customer Journey Mapping
• Service Blueprinting
• SIPOC
• Value Stream Mapping
• BPMN
• Process Mining
• Voice of the Customer
• Five Whys
• Fishbone Analysis
• Problem Tree
• Opportunity Solution Tree
• Assumption Mapping
• Stakeholder Mapping

Practical discovery sequence​

A strong discovery sequence is:

1. Stakeholder Mapping
2. Voice of the Customer
3. Jobs to Be Done
4. Customer Journey Mapping
5. Service Blueprinting
6. SIPOC
7. BPMN or Process Mining
8. Five Whys or Fishbone Analysis
9. Opportunity Solution Tree
10. Assumption Mapping

What each level examines​

Lens	Question it answers
Jobs to Be Done	What are users trying to accomplish?
Customer Journey	What do users experience?
Service Blueprint	What people and systems support the journey?
SIPOC	What are the process boundaries?
BPMN	What is the detailed workflow?
Process Mining	What actually happens in operational systems?
Five Whys / Fishbone	Why does the problem occur?
Opportunity Solution Tree	Which opportunities and solutions should be tested?

Discovery ends when you can state the problem, the evidence, the root causes and the assumptions that remain untested—not when you have filled every canvas.

---

4. Assess AI readiness and maturity​

Use these assessments to determine whether the organisation can successfully deliver and sustain AI. Enthusiasm is not readiness.

Essential assessments​

• Enterprise AI Readiness Assessment
• AI Capability Maturity Model
• Data Maturity Assessment
• AI Data Readiness Assessment
• Cloud Maturity Assessment
• Cybersecurity Maturity Assessment
• Responsible AI Maturity Model
• AI Governance Maturity Model
• MLOps Maturity Model
• LLMOps Maturity Model
• Agentic AI Readiness Assessment
• Change Readiness Assessment
• Operational Readiness Assessment

Recommended readiness dimensions​

A consolidated AI readiness assessment should cover:

1. Strategy and leadership
2. Use-case portfolio
3. Data
4. Technology and cloud
5. Architecture
6. AI engineering
7. MLOps, LLMOps and AgentOps
8. Security and privacy
9. Responsible AI and governance
10. People and skills
11. Change and adoption
12. Operating model
13. Vendor and ecosystem capability
14. Benefits and performance measurement

Use maturity models carefully​

Do not run separate, disconnected maturity exercises for every technical domain. Use one integrated maturity assessment, supported by specialised modules where needed. A fifteen-slide “maturity theatre” for each silo creates noise without a remediation backlog.

Output that matters: a heat map, a remediation backlog with owners, and a clear statement of which use cases are blocked until capability gaps close.

---

5. Prioritise AI use cases​

Use these frameworks to compare opportunities consistently and prevent the portfolio from being driven by enthusiasm alone.

Essential frameworks​

• Desirability–Viability–Feasibility
• Value–Feasibility–Risk
• Weighted Scoring Model
• Use-Case Portfolio Matrix
• Impact–Effort Matrix
• Strategic Alignment Scoring
• Risk-Adjusted Value Scoring
• Cost-of-Delay Analysis
• WSJF
• RICE
• Dependency-Aware Prioritisation
• Three Horizons Portfolio

Recommended enterprise prioritisation model​

Use a weighted score containing:

• Strategic alignment
• Customer or employee value
• Financial value
• Process impact
• Data readiness
• Technical feasibility
• Responsible AI risk
• Security and privacy risk
• Regulatory complexity
• Change complexity
• Time to value
• Reusability
• Scalability
• Delivery cost
• Operational sustainability

Recommended visualisation​

Place scored use cases on a portfolio matrix:

Quadrant	Meaning
Quick wins	High value, high feasibility
Strategic bets	High value, lower feasibility
Enablers	Lower direct value but reusable
Avoid or defer	Low value or unacceptable risk

Avoid overusing scoring methods​

Choose one primary scoring system:

• Weighted Scoring Model for enterprise portfolios
• RICE for product features
• WSJF for delivery sequencing
• Impact–Effort for workshops

Do not combine all scoring formulas into one unnecessarily complex model. Complexity in prioritisation is often a substitute for sponsorship clarity.

---

6. Build the commercial and value case​

Use these frameworks to prove that the initiative is worth funding—and to define who owns benefits after go-live.

Essential frameworks​

• Business Case
• Five Case Model
• Total Cost of Ownership
• Return on Investment
• Net Present Value
• Payback Period
• Break-Even Analysis
• Sensitivity Analysis
• Scenario Analysis
• Unit Economics
• Benefits Dependency Network
• Benefits Realisation Plan
• Value-Driver Tree
• Risk-Adjusted NPV
• Real Options Analysis

Core business-case structure​

A strong AI business case should cover:

1. Strategic case — why intervention is necessary
2. Economic case — which option provides the greatest net value
3. Commercial case — how technology and suppliers will be acquired
4. Financial case — whether the organisation can afford it
5. Management case — whether it can be delivered and governed

AI-specific cost categories​

TCO should include:

• Data acquisition and preparation
• Cloud infrastructure
• Model usage
• Token or inference consumption
• Vector databases
• Integration
• Licences
• Evaluation
• Security controls
• Human oversight
• Monitoring
• Support
• Model updates
• Vendor management
• Compliance and assurance
• Change and training
• Exit and migration costs

Benefits frameworks​

The Benefits Dependency Network is particularly useful because it links:

Technology capability → business change → operational outcome → measurable benefit

This prevents organisations from treating model deployment itself as a business benefit. If the network cannot name the business change and the owner of the benefit, the case is not ready for investment approval.

---

7. Design the operating model​

Use these frameworks to define how the organisation will own, govern and operate AI. Architecture without ownership produces shadow AI and unowned incidents.

Essential frameworks​

• Target Operating Model
• Operating Model Canvas
• AI Operating Model
• Product Operating Model
• Hub-and-Spoke Model
• Centre of Excellence
• Centre for Enablement
• Team Topologies
• Three Lines Model
• Galbraith Star Model
• Process Ownership Model
• AI System Ownership Model
• Build–Buy–Partner Framework
• Skills and Competency Framework

Key operating-model decisions​

The organisation must decide:

• What is centralised?
• What is federated?
• Who owns AI products?
• Who owns models, prompts and agents?
• Who approves high-risk use cases?
• Who monitors deployed systems?
• Who handles incidents?
• Who funds shared platforms?
• Who owns reusable components?
• Who manages vendors?
• Who owns benefits after deployment?

Recommended structure​

A common model is:

• Central AI enablement function — platforms, standards, reusable patterns and specialist expertise
• Federated business teams — use-case ownership, domain knowledge and adoption
• Independent risk functions — challenge, oversight and assurance
• Executive governance — investment, risk appetite and strategic direction

Operating-model design is complete when every material system has a named owner, every high-risk decision has a RAPID path, and platform funding is not an annual surprise.

---

8. Design enterprise and solution architecture​

Use these frameworks to create a coherent, secure and operable AI solution—not a slideware landscape.

Essential frameworks​

• TOGAF
• ArchiMate
• C4 Model
• Architecture Decision Records
• Domain-Driven Design
• Event Storming
• API-First Architecture
• Event-Driven Architecture
• Cloud Adoption Framework
• Well-Architected Framework
• Zero Trust Architecture
• Data Mesh
• Data Fabric
• Lakehouse Architecture
• Medallion Architecture
• Platform Engineering
• Team Topologies
• Architecture Trade-Off Analysis Method
• Quality Attribute Workshop

Recommended use by level​

Enterprise level​

• TOGAF
• ArchiMate
• Capability-Based Planning
• Target Architecture
• Transition Architecture

Solution level​

• C4 Model
• Domain-Driven Design
• API-First Architecture
• Event-Driven Architecture
• Architecture Decision Records

Cloud and operational level​

• Cloud Adoption Framework
• Well-Architected Framework
• Zero Trust
• SRE
• FinOps
• DevSecOps

Practical architecture views​

Every AI solution should document:

1. Business context
2. Users and actors
3. System context
4. Containers and services
5. Data flows
6. Model interactions
7. Agent and tool interactions
8. Identity and access
9. Security boundaries
10. Deployment architecture
11. Monitoring architecture
12. Failure and recovery mechanisms
13. Human oversight
14. Supplier dependencies

If a view is missing—especially human oversight, failure modes or supplier dependencies—treat the design package as incomplete at the design gate.

---

9. Govern data​

Use these frameworks to ensure data is usable, controlled and traceable. AI systems inherit the quality, legality and lineage of their data.

Essential frameworks​

• DAMA-DMBOK
• Data Governance Operating Model
• Data Ownership Model
• Data Stewardship Model
• Data Product Model
• Data-as-a-Product
• Data Contracts
• Data Quality Framework
• Data Lineage
• Metadata Management
• Data Classification
• Data Lifecycle Management
• Data Minimisation
• Data Observability
• Data Mesh
• Data Fabric
• Medallion Architecture
• AI Data Readiness Framework

Core data questions​

• Is the data available?
• Is it legally usable?
• Is it representative?
• Is it sufficiently accurate?
• Who owns it?
• How fresh is it?
• Can its origin be traced?
• Can changes be detected?
• Is sensitive data identified?
• Are retention rules defined?
• Can the organisation reproduce model outputs?
• Are retrieval sources authoritative?

Until these questions have named owners and evidence, model experimentation should remain constrained. Data unreadiness is a portfolio decision, not only a technical inconvenience.

---

10. Engineer AI systems​

Use these frameworks to structure the technical lifecycle. Pattern choice matters, but lifecycle discipline matters more.

Essential frameworks​

• CRISP-DM
• Team Data Science Process
• Machine Learning Lifecycle
• Agent Development Lifecycle
• MLOps
• LLMOps
• GenAIOps
• AgentOps
• DataOps
• EvaluationOps
• Human-in-the-Loop
• Retrieval-Augmented Generation
• Tool-Using Agent Patterns
• Durable Execution
• Model Context Protocol
• Experiment Tracking
• Model Registry
• Prompt Management
• Dataset Versioning

Recommended lifecycle​

1. Problem formulation
2. Data understanding
3. Data preparation
4. Baseline creation
5. Model or system selection
6. Prompt, retrieval or agent design
7. Experimentation
8. Evaluation
9. Risk and security testing
10. Human validation
11. Deployment
12. Monitoring
13. Incident management
14. Improvement or retraining
15. Retirement

Use the correct operational discipline​

Discipline	Use for
MLOps	Predictive and traditional machine-learning systems
LLMOps	Language-model applications
GenAIOps	Multimodal generative AI systems
AgentOps	Autonomous or semi-autonomous agent workflows
DataOps	Data pipelines and quality
EvaluationOps	Repeatable and continuous AI evaluation

Do not force every GenAI or agent system into a classic CRISP-DM checklist. Keep the problem-data-evaluate-deploy spine, then apply the operational discipline that matches the system type.

---

11. Evaluate and assure the AI system​

This is one of the most important additions to an end-to-end methodology. Without evaluation evidence, production is a bet.

Essential frameworks​

• AI Evaluation Framework
• LLM Evaluation Framework
• RAG Evaluation Framework
• Agent Evaluation Framework
• Golden Dataset
• Human Evaluation
• Automated Evaluation
• Offline Evaluation
• Online Evaluation
• Continuous Evaluation
• Champion–Challenger Testing
• Adversarial Evaluation
• Red Teaming
• Model Validation
• AI Assurance Case
• Production Readiness Review

Core evaluation dimensions​

• Correctness
• Relevance
• Groundedness
• Faithfulness
• Retrieval precision
• Retrieval recall
• Hallucination rate
• Task completion
• Tool-selection accuracy
• Agent trajectory quality
• Escalation accuracy
• Safety
• Bias and fairness
• Privacy
• Security
• Robustness
• Latency
• Reliability
• Cost
• User satisfaction
• Business outcome

Evaluation stages​

1. Component evaluation
2. End-to-end system evaluation
3. Adversarial evaluation
4. Human acceptance
5. Controlled pilot
6. Production monitoring
7. Regression evaluation
8. Periodic independent assurance

Evaluation is not a single pre-release event. It is a continuous control that feeds retirement as well as promotion. Golden datasets, red-team findings and production regressions should update the same evidence pack that gatekeepers review.

---

12. Manage responsible AI and regulatory risk​

Use these frameworks to make AI governance systematic and evidence-based—not a slide deck produced after the demo.

Essential frameworks​

• NIST AI Risk Management Framework
• ISO/IEC 42001
• ISO/IEC 23894
• ISO/IEC 42005
• EU AI Act Risk Classification
• Algorithmic Impact Assessment
• Responsible AI Impact Assessment
• AI System Inventory
• Model Cards
• Data Cards
• System Cards
• Responsible AI Control Library
• Human Oversight Framework
• Three Lines Model
• AI Assurance Case
• AI Incident Management
• AI Decommissioning Framework

Core responsible AI principles​

• Accountability
• Transparency
• Explainability
• Fairness
• Safety
• Robustness
• Privacy
• Security
• Human agency
• Contestability
• Traceability
• Auditability

Minimum governance artefacts​

Every material AI system should have:

• Named system owner
• Defined intended purpose
• Prohibited uses
• Risk classification
• Data and model documentation
• Evaluation evidence
• Human oversight design
• Control mapping
• Monitoring thresholds
• Incident route
• Change history
• Retirement criteria

If any of these are missing for a high-risk system, treat release as blocked regardless of model quality metrics.

---

13. Protect security and privacy​

Use these frameworks to protect data, models, tools, identities and third parties. AI expands the attack surface; it does not replace traditional security discipline.

Essential frameworks​

• NIST Cybersecurity Framework
• ISO/IEC 27001
• Zero Trust
• STRIDE
• MITRE ATT&CK
• MITRE ATLAS
• OWASP Top 10 for LLM Applications
• Privacy by Design
• Data Protection Impact Assessment
• Secure Development Lifecycle
• Defence in Depth
• Identity and Access Management
• Least Privilege
• Shared Responsibility Model
• Software Bill of Materials
• AI Bill of Materials
• Third-Party Risk Management
• Incident Response Framework
• Business Continuity
• Operational Resilience

AI-specific security areas​

• Prompt injection
• Indirect prompt injection
• Sensitive information disclosure
• Insecure tool use
• Excessive agent permissions
• Data poisoning
• Model extraction
• Membership inference
• Model inversion
• Supply-chain compromise
• Unsafe output handling
• Vector-store poisoning
• Cross-tenant data leakage
• Malicious documents
• Uncontrolled autonomous actions

Security work for AI must be threat-modelled early, tested adversarially before scale, and monitored continuously after release. A DPIA and an AI Bill of Materials belong in the same gate pack as the architecture decision records.

---

14. Deliver, release and scale​

Use these frameworks to organise experimentation, development and release without confusing a demo with a product.

Essential frameworks​

• Agile
• Scrum
• Kanban
• Dual-Track Agile
• Stage-Gate
• Lean Startup
• Product Operating Model
• DevOps
• DevSecOps
• Continuous Discovery
• Continuous Delivery
• Hypothesis-Driven Development
• Test-Driven Development
• Model-Driven Experimentation
• Proof of Concept
• Minimum Viable Product
• Pilot
• Progressive Delivery
• Production Readiness Review
• Operational Readiness Review

Recommended delivery progression​

1. Discovery
2. Feasibility experiment
3. Proof of concept
4. Prototype
5. Minimum viable product
6. Controlled pilot
7. Limited production
8. Progressive rollout
9. Enterprise scale
10. Continuous optimisation

Stage gates​

At each stage, decide whether to:

• Continue
• Redesign
• Expand
• Restrict
• Pause
• Stop

Stopping is a successful governance outcome when evidence shows unacceptable risk, weak value or unreadiness. Dual-Track Agile keeps discovery evidence flowing while delivery hardens what has already been validated.

---

15. Operate and monitor​

Use these frameworks after the system enters production. Production is where most AI value—and most AI harm—is realised.

Essential frameworks​

• ITIL 4
• Site Reliability Engineering
• Service-Level Indicators
• Service-Level Objectives
• Error Budgets
• Observability
• OpenTelemetry
• Model Monitoring
• LLM Observability
• Agent Observability
• Data Drift Monitoring
• Concept Drift Monitoring
• AI Incident Management
• Problem Management
• Change Enablement
• FinOps
• Business Continuity
• Disaster Recovery
• Operational Resilience
• Blameless Postmortems

What to monitor​

• Availability
• Latency
• Error rates
• Model quality
• Retrieval quality
• Agent completion
• Tool failures
• Escalations
• Policy violations
• Security events
• Data drift
• Model drift
• Cost per interaction
• Token consumption
• Human overrides
• Customer satisfaction
• Business benefits

Operations for AI must combine classic reliability practice with AI-specific quality signals. An SLO on latency alone is insufficient if groundedness collapses or human override rates spike.

---

16. Drive change and adoption​

Use these frameworks to ensure that people use the system correctly and consistently. Unused AI is an expensive experiment, not a transformation.

Essential frameworks​

• Prosci ADKAR
• Kotter’s Eight-Step Model
• McKinsey 7S
• Change Impact Assessment
• Change Readiness Assessment
• Stakeholder Influence–Interest Matrix
• Communications Planning
• Training-Needs Analysis
• Technology Acceptance Model
• COM-B
• Behaviour Change Wheel
• Communities of Practice
• Champion Networks
• AI Ambassador Programme
• 70–20–10 Learning Model
• Kirkpatrick Evaluation Model
• Human–AI Collaboration Framework
• Adoption Measurement Framework

Recommended combination​

• McKinsey 7S — assess organisational alignment
• Change Impact Assessment — identify who and what will change
• ADKAR — manage individual adoption
• Kotter — manage broader organisational mobilisation
• COM-B — diagnose behavioural barriers
• Champion Network — provide local support and reinforcement
• Kirkpatrick — evaluate training effectiveness

Adoption measurement should track behaviour and outcome quality, not only licence counts or login volume. Human–AI collaboration design belongs in the operating model, not only in the training pack.

---

17. Manage vendors and procurement​

Use these frameworks when external platforms, models or implementation partners are involved. Supplier choices create long-lived architectural and regulatory commitments.

Essential frameworks​

• Build–Buy–Partner Framework
• Make-or-Buy Analysis
• Vendor Evaluation Matrix
• Request for Information
• Request for Proposal
• Proof of Value
• Technical Due Diligence
• Security Due Diligence
• Privacy Due Diligence
• Responsible AI Due Diligence
• Third-Party Risk Management
• Concentration Risk Assessment
• Vendor Lock-In Assessment
• Exit Management Plan
• Service-Level Agreement
• Shared Responsibility Model
• AI Bill of Materials

Important supplier criteria​

• Model capability
• Security
• Privacy
• Data residency
• Regulatory support
• Explainability
• Availability
• Scalability
• Integration
• Portability
• Pricing predictability
• Model deprecation policy
• Incident notification
• Audit rights
• Subprocessor controls
• Exit support

A Proof of Value without exit planning, subprocessors clarity and Responsible AI due diligence is incomplete commercial diligence. Treat concentration risk and lock-in as first-class portfolio risks.

---

18. Measure performance and benefits​

Use these frameworks to prove that the solution continues to produce value after the launch announcement fades.

Essential frameworks​

• Objectives and Key Results
• Balanced Scorecard
• North Star Metric
• Key Performance Indicators
• Key Risk Indicators
• Benefits Realisation Management
• Benefits Dependency Network
• Value-Driver Tree
• DORA Metrics
• Flow Metrics
• Customer Satisfaction Score
• Customer Effort Score
• Adoption Rate
• Task Completion Rate
• Human Override Rate
• Cost per Successful Outcome

Measurement layers​

Model metrics​

• Accuracy
• Precision
• Recall
• Groundedness
• Hallucination
• Robustness

System metrics​

• Latency
• Availability
• Failure rate
• Completion rate
• Escalation rate
• Cost

User metrics​

• Adoption
• Satisfaction
• Trust
• Effort
• Retention

Business metrics​

• Revenue
• Cost reduction
• Productivity
• Risk reduction
• Cycle-time reduction
• Service quality
• Employee capacity

Benefits realisation closes the loop with the commercial case and the operating model. If no one owns the benefit, and no operational change was delivered, model metrics alone will not save the investment thesis.

---

Final recommended core library​

For a practical ConsultAI Lab, the following approximately fifty frameworks provide sufficient end-to-end coverage. Keep the larger catalogue as a reference library. Run this set as the default methodology.

Mobilisation​

• Project Charter
• RACI
• RAPID
• RAID Log
• Stakeholder Influence–Interest Matrix
• MECE Issue Tree

Strategy​

• Strategy Choice Cascade
• Three Horizons
• Value Chain Analysis
• Business Model Canvas
• Operating Model Canvas
• Capability-Based Planning
• Wardley Mapping
• Value-Driver Tree

Discovery​

• Design Thinking
• Double Diamond
• Jobs to Be Done
• Customer Journey Mapping
• Service Blueprinting
• SIPOC
• BPMN
• Process Mining
• Five Whys
• Fishbone Analysis
• Opportunity Solution Tree

Readiness and prioritisation​

• AI Capability Maturity Model
• Data Maturity Assessment
• Responsible AI Maturity Model
• Agentic AI Readiness Assessment
• Desirability–Viability–Feasibility
• Value–Feasibility–Risk
• Weighted Scoring Model
• Use-Case Portfolio Matrix

Commercial​

• Five Case Model
• TCO
• ROI
• NPV
• Sensitivity Analysis
• Benefits Dependency Network
• Benefits Realisation Plan

Architecture and engineering​

• TOGAF
• C4 Model
• Domain-Driven Design
• Cloud Adoption Framework
• Well-Architected Framework
• Zero Trust
• Medallion Architecture
• MLOps
• LLMOps
• AgentOps
• DevSecOps
• SRE
• FinOps

Evaluation and governance​

• AI Evaluation Framework
• RAG Evaluation Framework
• Agent Evaluation Framework
• Golden Dataset
• NIST AI RMF
• ISO/IEC 42001
• EU AI Act Risk Classification
• Algorithmic Impact Assessment
• Model Cards
• Data Cards
• Human Oversight Framework

Security​

• NIST Cybersecurity Framework
• STRIDE
• MITRE ATLAS
• OWASP Top 10 for LLM Applications
• Privacy by Design
• DPIA
• Secure Development Lifecycle

Delivery, adoption and operations​

• Dual-Track Agile
• Stage-Gate
• Product Operating Model
• Continuous Delivery
• Production Readiness Review
• ITIL 4
• ADKAR
• Kotter Eight-Step
• McKinsey 7S
• Champion Networks
• Technology Acceptance Model
• Benefits Realisation Management

---

How to use the core set in practice​

1. Start from the spine. Map the engagement to 8D, VALUE and the relevant lifecycle (system, agent, responsible AI, secure development, benefits).
2. Pick one primary framework per decision. Use RAPID or RACI, not both for the same decision type; use one scoring model, not four.
3. Sequence discovery before prioritisation. Evidence first, enthusiasm second.
4. Integrate readiness. One assessment with specialised modules beats a maturity model zoo.
5. Force commercial honesty. Include AI-specific TCO and a Benefits Dependency Network before scale funding.
6. Design ownership with architecture. Operating model and C4 views travel together.
7. Treat evaluation and Responsible AI as release controls. Golden datasets, impact assessments and human oversight are gate evidence.
8. Measure what the business funded. Model metrics without business outcomes are incomplete.

This is the most useful core framework set. The larger catalogue can remain as a reference library. These frameworks form the runnable end-to-end AI solution-engineering methodology for ConsultAI Lab.