AI CAPABILITY • FOUNDATION
Data & Knowledge Foundation
The infrastructure AI needs to work
In 30 Seconds
AI implementation depends on three pillars: data quality, knowledge architecture, and technical infrastructure. Most organisations focus on the technology and underinvest in the first two.
The uncomfortable truth: Many organisations discover their knowledge gaps only when they try to use AI. Policies aren't as clear as assumed. Processes aren't as documented as believed. AI exposes what was hidden.
Our view: Knowledge is more valuable than data alone. The organisations getting results from AI aren't necessarily those with the most data — they're those who've structured their knowledge so AI can use it.
Knowledge > Data
Everyone talks about “data readiness” for AI. But there's something more fundamental that most organisations miss: knowledge readiness.
The Chatbot Test
Put a chatbot in front of your internal documentation. Ask it about your policies, procedures, how things work.
Most organisations discover: “Wait — we don't actually know what our policies are. They're not as clear as we thought.”
AI doesn't create this problem — it reveals it. The knowledge gaps were always there. They just weren't visible until you tried to use AI.
Data Architecture
Where your data lives, how it's structured, who can access it.
- • Databases, warehouses, lakes
- • Schema and structure
- • Access and governance
- • Quality and completeness
Knowledge Architecture
What your organisation actually knows — and whether it's accessible.
- • Documented policies and procedures
- • Institutional know-how
- • Decision frameworks
- • Domain expertise (captured or not)
The distinction matters: You can have excellent data infrastructure and still have AI that doesn't work — because the knowledge layer is missing. AI needs to know what things mean, not just where data lives.
Data Readiness
The foundation layer — necessary but not sufficient
Data Quality
AI is ruthless at exposing data quality issues. Inconsistencies, gaps, and errors that humans work around become blockers for AI systems.
“GenAI has been fantastic at elevating awareness of the necessity for good data.”
Data Governance
Who owns what data? Who can access it? What are the business definitions? AI projects stall when these questions don't have clear answers.
“Data is a team sport.”
RAG Readiness
Retrieval-Augmented Generation requires data to be structured for AI consumption. Metadata, chunking, embedding strategies all need to be considered.
Data needs to be “AI-ready”, not just stored.
Three Types of Metadata
Most organisations have the first. Few have all three:
Technical Metadata
Schema, structure, formats — where data lives
Business Metadata
Definitions, glossary, business context — what data means
Social Metadata
Who uses it, how, popularity — often missing entirely
What Data Can Go Where?
The question every organisation asks: “Can we put our data into AI tools?”
The real question: Not whether to use AI, but which AI, with which data, under what controls. The answer depends on your data classification, risk tolerance, and regulatory context.
The Options Spectrum
AI deployment options range from convenient (less control) to secure (more setup):
Public AI (Free Tiers)
CautionChatGPT free, Claude free, etc. Data may be used for model training. Generally unsuitable for confidential business data. Fine for public information and personal learning.
Enterprise Tiers
Check TermsChatGPT Enterprise, Claude for Enterprise, etc. Typically no training on your data, better retention policies, admin controls. Read the terms carefully — they vary.
API Access
More ControlDirect API access (OpenAI, Anthropic, etc.). Data typically not used for training. You control what's sent, logged, and retained. Requires technical integration.
Private / Local Deployment
Full ControlRun models locally (Ollama) or in your own cloud. Data never leaves your environment. Maximum security, but requires technical capability and accepts smaller models.
The Policy Layer
Data Classification
What's public, internal, confidential, restricted? Most organisations have this for traditional systems — it needs extending to AI tools.
Tool Approval Matrix
Which AI tools are approved for which data classifications? Who decides? Clear policy prevents both over-caution and risky shortcuts.
Where we help: Navigating this landscape, understanding the trade-offs, and building policies that enable AI adoption without unacceptable risk. The goal is informed choice, not blanket prohibition.
Technical Infrastructure
The technical layer that enables AI — but doesn't guarantee it works.
Infrastructure
- • Cloud architecture and deployment
- • Model hosting and inference
- • Scalability and performance
- • Cost optimisation
- • Monitoring and observability
Security & Integration
- • Data security and privacy
- • API design and management
- • Enterprise system integration
- • Authentication and access control
- • Compliance and audit trails
Our position: We partner with technical specialists for enterprise-scale infrastructure. Our focus is the layer above — ensuring technical foundations translate into working AI capability through context, knowledge, and fluency.
Right-Sized Solutions
Not every AI implementation needs enterprise infrastructure. For many organisations, the right approach is working with existing platforms and focusing investment on knowledge and capability — not custom technical builds.
Common Patterns We See
AI Aspiration, Missing Basics
Organisations excited about AI but lacking fundamentals: no data stewards, no catalog, unclear ownership. They want to run before they can walk.
Our approach: Build the missing foundations in parallel with early AI pilots.
Technology Over-Indexing
“Every technology under the sun” but 2-3 people using each. Tools without adoption. The presence of technology doesn't equal success — adoption does.
Our approach: Focus on making existing investments work before adding more.
Executive vs Working-Level Gap
Senior leaders think data and AI readiness is much higher than working-level staff report. This perception gap blocks progress.
Our approach: Surface reality through structured assessment, then align around truth.
Deep Dive: Data Capability
Data readiness is a foundation for AI — but it's also a capability in its own right. Data strategy, governance, and architecture matter beyond AI applications.
Explore Data Capability →From Foundation to Practice
Data and knowledge foundations enable the practice layer — where AI actually creates value.
Context Engineering
Knowledge architecture comes to life through context — the right information at the right time.
Learn more →Agents & Orchestration
Skills-based systems that build on your knowledge foundation to deliver real capability.
Learn more →AI Skills & Fluency
The human capability that turns foundation and practice investments into real value.
Learn more →Not Sure Where to Start?
Many AI initiatives stall because the foundations weren't in place. If you're not sure whether your data and knowledge are AI-ready — or you're discovering gaps as you go — let's talk.
We can help you understand what's really needed before you invest further.