Data and Knowledge

Most AI projects fail at the input stage. Half the business runs in the founder's head, the other half lives in 14 spreadsheets nobody updates, and the model never gets the chance to do anything useful.

This section is about getting the inputs right. The data that actually exists, the knowledge that lives between people's ears, and the filing system that nobody has audited since 2019. Practical work to get a small business AI-ready, with a 90-day path through the bits that matter and the things you can park.

Two people at a conference table, one looking at a laptop screen with a concerned expression

Find the shadow AI in your agency before a client's data leaks through it

Agency teams paste client material into free AI tools daily, often without data agreements, and that exposure may already be a breach of your client confidentiality contract.

Two colleagues in a meeting room reviewing a printed data classification table on a conference table

A four-tier data map so your team knows what AI can touch

A single-page data classification table that maps four data tiers to the AI tools allowed to process each one, so every team member knows the rule before they paste.

An experienced factory worker showing a younger colleague how a piece of equipment works at a workshop bench

Capture the shop-floor knowledge before it retires

The real asset in a specialist manufacturer is rarely on the balance sheet. It lives in the heads of your most experienced operators, and when they retire, it goes with them unless someone has captured it.

Person reviewing paper records at a workshop bench with industrial machinery in the background

Why AI stalls in manufacturing when the data lives on paper

In specialist manufacturing, the data that AI tools need is often on paper or trapped in systems that have never been connected, and that is what stalls the programme, not the technology.

Two people at an office desk looking at a laptop screen, one pointing at it while the other reads a printout

Your AI is only as good as your data

When an AI pilot keeps producing confident answers your team quietly knows are wrong, the model is rarely the problem. The data underneath it is.

A person at a laptop in a small office, reviewing a folder structure on screen

Best-practice folder and sharing structure for Google Drive

The five-drive Google Drive structure every UK services firm needs: Shared Drives by function, consistent naming, and compliant access control.

two business professionals reviewing documents together at a desk in a naturally lit office

Australian data sovereignty explained for UK business owners

Australian data sovereignty law applies to UK businesses that handle personal data from Australian clients. Here is what it covers, where you will encounter it, and what to do first.

A person sitting at a desk with a laptop open in a tidy office, with neatly arranged papers and a coffee cup nearby

A straightforward way to organise Google Drive at work

A practical folder structure and setup guide for owner-managed UK services firms getting Google Drive properly under control.

Person at a desk reviewing paper documents alongside a laptop in a small office

OCR versus text extraction: what each one is good for

The difference between OCR and text extraction determines whether your document automation saves time or creates new checking work.

person reviewing printed documents at a desk with a laptop open beside them

Data management basics for owner-managed firms

A plain-English guide to data management for owner-managed firms: what it means in practice, why UK GDPR makes it a legal obligation, and how the basics make AI adoption safer and more effective.

Business owner reviewing documents at a desk with a laptop open and natural light from a window

What data minimisation means under GDPR, in plain English

Data minimisation is one of GDPR's core principles and one of the least understood, so here is what it actually requires from an owner-managed UK service firm, including when AI tools change the calculation.

Person at a desk reviewing handwritten notes beside an open laptop

What a knowledge base system is and when it helps

A knowledge base system can cut your team's search time and reduce repeated questions, but only when your processes are already repeatable and someone will maintain it.

A person at a desk reviewing files on a laptop with physical folders arranged nearby

How storage systems save space by removing repeats

Data deduplication keeps one copy of repeated data and references the rest, reducing backup costs on workloads that have genuinely repetitive content.

A business owner at a desk reviewing files on a laptop with paper documents nearby

A plain-English explanation of data deduplication

What data deduplication actually means, why UK GDPR's data minimisation rules make it relevant, and when it is worth acting on for owner-managed service businesses.

Person at a desk reviewing a printed document alongside an open laptop in a naturally lit office

Where to store SOPs so staff can find and trust them

A practical decision guide for owner-managed businesses on choosing where to store SOPs so staff can find and trust them.

Two people at a desk reviewing printed documents spread between them

Simple naming conventions for files, folders and records

A documented naming convention tells your team what to call every file before saving it, so records stay findable, recoverable, and compliant with ICO obligations from day one.

Person at an office desk scrolling through a laptop with a notebook open beside them

Make Slack and email knowledge easier to search and reuse

Simple conventions cut Slack and email search time significantly, and UK services firms face a compliance dimension worth understanding.

Person sitting at a desk reviewing a spreadsheet on a laptop with handwritten notes beside them

A practical approach to cleaning messy business data

A practical, sequenced guide to cleaning messy data in a 5-50 person services firm, with UK regulatory context and a governance checklist to stop it recurring.

Two colleagues reviewing a document structure on a laptop at an office table

What a small business knowledge management system should include

A practical guide for owner-managed businesses on when a knowledge management system is worth the investment, what features actually drive adoption, and what questions to ask before you sign anything.

A business owner at a desk reviewing documents on a laptop with a folder of papers nearby

RAG versus large context windows: which suits your business?

Know whether RAG or a large context window fits your business's documents before your vendor decides for you.

A business owner sitting at a desk reviewing data on a laptop with printed documents beside them

Practical rules for deduplicating data without causing harm

A practical playbook for cleaning duplicate records in your CRM and marketing data, safely, without deleting consent history or triggering a UK GDPR breach.

Person reviewing a structured spreadsheet on a laptop at a well-lit desk near a window

How to keep a data dictionary accurate, useful, and current

For a services firm with 5 to 50 staff, a data dictionary that nobody maintains is just documentation debt. Here is how to keep yours accurate, useful, and current.

Two professionals reviewing documents at an office desk, one gesturing towards a laptop screen

Data classification in regulated financial services firms

Why data classification is the organising principle behind every security, compliance, and AI decision in a regulated UK financial services firm.

Two colleagues reviewing documents on a laptop in a naturally lit office

RAG or long context: how to choose the right architecture for your business AI

A practical decision guide to choosing between retrieval-augmented generation and long-context prompting for business AI, covering when each fits, what it costs to get wrong, and the questions to ask before committing.

Person at a laptop reviewing a tidy document folder structure in a well-lit office

A Google Drive folder template for small businesses

A five-drive folder structure for owner-managed businesses on Google Workspace, covering client work, operations, HR, and UK GDPR compliance without overcomplicating things.

Two colleagues reviewing a shared spreadsheet at a conference table, one pointing at the screen

Why every small business needs a simple data glossary

Finance says 240 customers. Marketing says 340. A data glossary is how owner-managed businesses ensure every team means the same thing by the same word.

Two business owners reviewing documents together at a desk, one looking at a laptop and the other holding a printed folder.

When centralising company documents is worth the effort

A decision guide for UK SME founders weighing whether to centralise company documents, when it pays off, when it does not, and what to ask before deciding.

Two people sitting at an office desk looking at a laptop screen together in natural daylight.

What deduplication is and how systems remove repeated data

Plain-English explainer for UK service-firm owners on what data deduplication actually is, where it pays back, and what to ask your backup or CRM supplier.

A business professional reviewing printed documents at a desk in a well-lit office

What is data classification? Why it matters for your business

How labelling your data by sensitivity gives AI tools, copilots, and automation the instruction they need, and why UK regulators now assume the work has been done.

Person reviewing spreadsheet data on a laptop at a tidy office desk

Metadata versus master data: a plain-English guide for small firms

Many owner-managed firms have both metadata and master data problems without realising it. Here is what each one is, why the distinction matters, and what good practice looks like without a data team.

Business owner reviewing documents at a desk with a laptop open beside them

How RAG uses your company documents to answer questions well

Retrieval-augmented generation connects an AI to your own documents so it can answer questions from them directly, here is how the pipeline works and what it means for your firm.

Founder at a desk reviewing printed checklists with a laptop open to a documentation workspace

What a small business knowledge base is and why it matters

What a knowledge base actually is for a small UK service firm, why it matters for time and key-person risk, when it is worth building, and what data protection rules apply before you start.

Person at a desk writing notes in a notebook while reviewing content on a laptop

What is a knowledge base? A plain-English guide for UK service firms

A plain-English definition of a knowledge base, the common uses for small UK service firms, and the compliance basics you need to know.

A person sitting at a desk reviewing documents with a laptop open beside them

Is knowledge management software worth the cost for your SME?

A practical decision guide for UK SME founders on whether knowledge management software is worth the cost, and what to ask before you commit.

Person sitting at a desk reviewing documents on a laptop in a well-lit office

When SharePoint works for SME knowledge management, and when it doesn't

SharePoint is already included in your Microsoft 365 licence, but whether it's the right knowledge management system for your firm depends on a question many founders haven't answered yet.

A person sitting at a desk in a natural-light office, reviewing data on a laptop screen

Preventing duplicate customer records in everyday systems

A practical guide to preventing, detecting and resolving duplicate customer records for UK owner-operated services firms.

Person at a desk reviewing documents in a naturally lit office

Right-sized data governance for small teams

A practical guide to data governance for UK small businesses: the four policies, three roles, and crown-jewel approach that cover what the ICO requires without building enterprise-scale bureaucracy.

Two people at a desk reviewing a printed document together in an office

How to get staff to document processes in a usable, repeatable way

A practical guide to getting process documentation out of shared drives and into daily use in owner-managed UK services firms.

Two colleagues reviewing a shared document on a computer screen in a small office

Build a business glossary for consistent terms and reporting

If your CRM, accounting system, and project tool all use your key business terms differently, a business glossary is the practical fix.

A person reviewing printed documents and notes at a desk with an open laptop, working in a well-lit office.

RAG vs long context: how to choose the right approach for your business

RAG retrieves what the model needs; long context loads more of it directly. Here is how to decide which approach fits your knowledge base, and what each costs to get wrong.

Person reviewing documents alongside an open laptop at a wooden desk

When to use RAG versus long-context models for SME knowledge work

A practical decision guide for UK SME founders on when to use retrieval-augmented generation versus long-context models for internal knowledge work.

Two colleagues reviewing documents at an office desk, one pointing to a laptop screen

When a knowledge base is better than a database

A practical decision guide for UK founders on when a knowledge base outperforms a database, when to choose differently, and what the wrong call typically costs.

Person at a desk reviewing documents beside an open laptop

How data minimisation and storage limitation differ in practice

Two distinct UK GDPR principles, data minimisation and storage limitation, apply at different stages and cover different decisions. Here is how to tell which one is in play.

A woman reviews a data spreadsheet on a laptop at a clinic desk

A healthcare data dictionary template for patient, clinic, and reporting data

Build a working data dictionary for your healthcare practice using NHS field standards, with guidance on UK GDPR compliance, AI governance, and commissioner reporting.

Two professionals reviewing documents and a laptop screen at an office desk

Data duplication versus redundancy in business data systems: a decision guide

Data duplication and data redundancy are opposite problems with similar-sounding names, and UK founders who conflate the two tend to fix the wrong one at real cost.

Business owner reviewing storage or backup data on a laptop screen at a desk in a modern office

Deduplication versus compression: what each one actually does

Deduplication and compression both reduce storage costs, but they solve different problems, perform differently by workload, and interact in ways that matter if you apply the wrong one.

Person reviewing documents and folders on a laptop at a tidy desk with natural light

File organisation rules that make AI search actually useful

The way you organise your files determines whether an AI search tool like Copilot saves your team time or creates a data protection problem.

Two colleagues having a working conversation across a desk, one taking handwritten notes

Capturing staff know-how before it walks out the door

A practical playbook for capturing the staff know-how that makes your business work, before the next key person hands in their notice.

Person at a desk reviewing documents on a laptop with a stack of printed files nearby

Why AI gives wrong answers from your company files

AI tools that read your company files can still give confident wrong answers. Here's why retrieval-augmented generation fails in practice and what a UK owner-managed firm should do first.

Person reviewing documents on a laptop at a desk

How to stop your AI from answering with out-of-date files

A practical guide to keeping AI tools away from stale files, outdated price lists, and superseded policies in a 5-50 person services firm.

A person reviewing printed documents at a desk with a laptop open nearby

How to prepare your business data for AI tools

A practical playbook for getting your business data AI-ready, addressing both the legal requirements and the common mistakes that make AI tools unreliable.

A person at a desk with folders of documents open beside a laptop

How AI handles stale documents, contradictions, and version drift

Version drift is the AI error that looks like a correct answer until a client challenge reveals the flaw. Here is what causes it and what to fix.

Person at a desk reviewing documents on a laptop screen with a stack of paper files beside them

How to delete old business files without breaking your records or search

A practical guide for owner-managed UK firms on safely deleting old files, protecting legal records, and keeping business search working.

Two colleagues looking at printed documents spread across an office desk

Data reduction versus deduplication: the SME decision guide

A plain-English decision guide for UK SMEs on when to prioritise deduplication over compression, when compression-first makes more sense, and five questions to ask before committing to a storage configuration.

Two colleagues reviewing a document together at an office desk

When an AI-powered knowledge base is worth it for your firm

The decision guide for UK SME owners on when an AI knowledge base pays off, when shared documents are enough, and what to ask before committing.

Two people sitting at a table together reviewing printed documents, one pointing to a page

Writing SOPs that AI systems can read and follow reliably

Learn how to structure your process documents so AI tools can actually follow them, with the data-handling rules UK regulators expect to see.

Two people at a small office desk reviewing documents and a laptop together by a window

Prompt injection and RAG safety, the new data risk for SMEs

Once you connect an AI assistant to your internal documents, a new class of attack opens up that firewalls do not catch. Here is what prompt injection actually is and the proportionate safeguards a small firm can put in place.

Two people reviewing printed documents and a laptop at a desk in an open-plan office, one holding a coffee.

RAG for SMEs in 2026: long-context models vs vector databases

Long-context models have changed the architecture call. Here is how a 5 to 50 person firm should choose between long prompts, lightweight RAG, and a full vector stack in 2026.

A practice manager at a small firm reviewing a printed data retention policy at her desk with a laptop open beside her

AI, GDPR and the EU AI Act: data retention rules for SMEs

Your AI assistant sits between two clocks. GDPR says delete. The EU AI Act says keep. Here is how a small firm reconciles them without retrofitting policy after the fact.

An SME owner at a desk reviewing a thirty-seven tab operations spreadsheet alongside a half-configured new system on a second screen

From spreadsheets to systems: when an SME outgrows its data setup

The spreadsheet was the right answer at five people. At twenty-five, it is the bottleneck. Four signs to watch for, and a proportionate migration sequence that does not retire the spreadsheet on day one.

An owner and two colleagues at a small meeting table with a printed sheet of questions, a coffee cup and an open notebook, mid-conversation

Measuring data and knowledge readiness, four questions to revisit each quarter

Owners who measure data and knowledge readiness in a structured way ship better AI outcomes than owners who treat it as a feeling. Four questions, every quarter, no dashboard required.

A woman in her fifties at a kitchen table reading a printed proposal document, pen in hand, with two other proposals beside her and a mug of tea.

When to bring in a data consultant, and when not to

Three situations where external data help is worth the spend, the more common ones where in-house is faster, and how to brief a consultant so the engagement actually lands.

A founder at a desk with a one-page plan in her hand, checking items off, with a three-block wall planner behind her

The 90-day data and knowledge readiness roadmap for SMEs

A focused ninety-day sprint, structured in three phases of about a month each, gets a 5 to 50 person business from scattered information to data and knowledge that does not block the next AI use case.

A small business owner and her operations lead at a desk, looking at a laptop screen showing a folder of documents, the owner pointing at one file

Version control for SME knowledge, the proportionate approach

Proposal_v4_dh_FINAL_final2.docx was tolerable when only humans read it. The moment AI is reading your documents the cost shows up sharply. Here are the three rules that do the work.

A small business owner and two senior colleagues around a meeting table, a printed glossary sheet in the centre, one writing on it with a pen, another pointing at a line

Building a business glossary, the foundation most SMEs skip

One page, twenty terms, ninety minutes. The cheapest and highest-impact AI readiness investment a small business can make, and the one almost nobody does.

A small business owner at a desk reviewing a printed sheet of tag categories alongside a CRM contact list on a laptop

Tagging and metadata: the SME search problem nobody named

Every SME has a metadata problem and almost none has named it as one. A proportionate fix is ten to fifteen tags, agreed once, reviewed quarterly.

Two colleagues at a desk reviewing a folder structure on a laptop with a printed three-column tier list on the desk

The shared drive problem, getting a decade of documents AI-ready

An owner's new AI tool surfaced a 2019 PDF as the answer to a current client question. Three more recent and relevant documents were sitting two folders away. The fix is not a clean-out, it is a categorisation pass.

A woman at a desk reviewing a shared-drive folder structure on her laptop with a printed checklist beside her

Naming conventions and folder structure for AI retrieval

Your AI assistant just cited Client_Proposal_FINAL_v3 (2).docx as the latest version of a document that was superseded three months ago. The fix is four rules, applied going forward, with a forgiving attitude to history.

An owner at a desk comparing two PDFs on a laptop screen with a folder of paper contracts open beside him

PDFs, scans and image-only documents, what AI can and cannot read

Two PDFs that look identical on screen can behave very differently when an AI tool tries to read them. Here is the difference, why it matters, and how to triage a mixed document estate without OCR-ing the lot.

An owner at a small desk with a laptop, notepad, and phone showing recent client calls, searching for an answer that is already in the archive

Slack, email and call transcripts: the hidden knowledge base sitting in your messages

Your team has already answered the question dozens of times. The answers are sitting in Slack, in old email threads, and in the call recordings nobody re-listens to. AI has finally made that archive useful, but only if you scope it proportionately.

A woman at a kitchen table reading and annotating a printed standard operating procedure with a pen

Standard operating procedures AI can actually read

An owner asks the AI to onboard a new client using the standard process. It invents half the steps. The fix is not a new tool, it is three small changes to how the SOP is written.

A senior operations lead in her late fifties points at a laptop screen while a younger colleague takes notes on a paper pad, the two of them at a small meeting table with a printed process diagram on the wall behind them

Capturing tacit knowledge before your key people leave

The most valuable knowledge in an owner-operated business is almost never written down. When a long-serving person leaves, the cost is measured in months of pattern-matching the replacement cannot do. Here is the proportionate capture discipline that actually works.

A small business owner and her operations lead at a kitchen table, two laptops open, a printed customer list between them, the owner pointing at a row on one screen

If any of this sounds familiar, let's talk.

The next step is a conversation. No pitch, no pressure. Just an honest discussion about where you are and whether I can help.

Book a conversation