Getting More From AI

General-purpose AI tools like ChatGPT and Claude are trained on public information. They know a lot about the world, but they know nothing about your business — your products, your customers, your internal processes, your pricing, your documentation.

This is the most common frustration people hit after the initial excitement of using AI wears off: "It gives good generic answers, but it doesn't know anything about us."

The good news is that there are now practical, accessible ways to give AI access to your business information — so it can answer questions, draft content, and support decisions based on your actual data, not just general knowledge.

The Simple Version: Uploading Documents

The most straightforward way to use your business data with AI is to upload documents directly into the conversation.

Most AI tools — including ChatGPT, Claude, and Gemini — now let you upload PDFs, Word documents, spreadsheets, and other files. The AI reads the document and can then answer questions about it, summarise it, or use it as context for whatever you ask next.

Examples of what this looks like in practice:

• Upload your employee handbook and ask: "What does our policy say about remote working on Fridays?"
• Upload a competitor's pricing PDF and ask: "How does this compare to our current pricing? Where are the biggest gaps?"
• Upload a long contract and ask: "Summarise the key obligations for both parties and flag anything unusual."
• Upload last quarter's sales spreadsheet and ask: "What are the top trends? Which product categories grew fastest?"

This works well for: One-off questions, ad-hoc analysis, and situations where you need the AI to reference a specific document.

The limitation: The AI only has access to what you upload in that session. Most AI tools don't remember your documents between conversations by default — though some now offer memory features or "projects" that retain context. And you still can't upload your entire company knowledge base into a single chat.

The Next Step: Knowledge Bases and RAG

When uploading individual documents isn't enough, the next step is a knowledge base — a collection of your business documents that an AI tool can search through and reference whenever it answers a question.

The technical term for this is RAG — Retrieval-Augmented Generation. It sounds complex, but the concept is straightforward:

1. You give the AI access to a collection of your documents (product guides, FAQs, process documents, etc.)
2. When someone asks a question, the AI first searches your documents for relevant information
3. It then uses what it finds to generate an accurate, specific answer — grounded in your actual business information rather than generic knowledge

Why this matters: It's the difference between an AI that gives you textbook answers and one that gives you answers based on how your business actually works.

Real examples:

• A customer service chatbot that answers questions using your actual product documentation and return policies — not generic responses
• An internal assistant that your team can ask questions about company processes, and it pulls the answer from your operations manual
• A sales tool that can reference your full product catalogue and pricing when helping draft proposals

How to get started with RAG:

Several tools now make this accessible without needing a developer:

For most small businesses, starting with custom GPTs or Claude Projects is the simplest first step. You upload your key documents, and the AI uses them to give more relevant, accurate answers.

Connecting AI Directly to Your Tools: MCP

There's a newer development worth knowing about, even if you're not ready to use it today: MCP — the Model Context Protocol.

Here's the idea in plain language: instead of uploading documents manually or setting up a knowledge base, MCP lets AI tools connect directly to your business software — your CRM, your database, your file storage, your project management tool — and pull in information as needed.

Think of it like giving the AI a set of keys to your filing cabinets, rather than having to photocopy documents and hand them over every time.

What this looks like in practice:

• You ask your AI assistant: "What's the status of the Johnson project?" — and it checks your project management tool directly
• You say: "Draft a follow-up email to the client we met last Tuesday" — and it pulls the meeting notes from your calendar and CRM
• You ask: "How much did we spend on marketing last quarter?" — and it queries your accounting software

Why MCP matters:

Before MCP, every AI tool needed its own custom connection to each piece of software. MCP creates a standard way for AI to connect to any tool — like how USB became a universal standard for connecting devices. This means:

• One protocol that works across many different AI tools and data sources
• Less custom development needed to connect your systems
• Your AI gets smarter because it can access real-time information, not just uploaded snapshots

Where it is today:

MCP is now supported by all the major AI providers and has become an industry-wide standard. For most small businesses, it's worth understanding the concept now so you can recognise the opportunity as the tools become more accessible. If you have a technical team or work with a developer, it's already worth exploring.

Keeping Your Data Safe

Using your business data with AI raises important questions about security and privacy. Here's what to keep in mind:

Understand where your data goes. When you upload a document to an AI tool, it's sent to that provider's servers. Read the data usage policy — most providers use data from consumer plans to improve their models, though you can usually opt out. Business plans typically have stronger protections.

Use business accounts, not personal ones. Most AI providers offer business or enterprise plans that include commitments around data handling, privacy, and not using your data for training. If you're putting business information into AI tools, use the appropriate plan.

Set clear team guidelines. Make sure your team knows which AI tools are approved, what types of information can be shared with them, and what's off-limits. This is especially important given the "shadow AI" problem — employees using AI tools without guidance.

→ For more on this topic, see Section 06 of the Business Guide: Security, Privacy & Risk

Once you're comfortable using AI for individual tasks, the next level is getting your tools to work together — so that actions in one system automatically trigger actions in another, with AI handling the parts that used to require a person.

This is where AI meets automation, and it's where many businesses see the biggest time savings.

What This Looks Like in Practice

The Tools That Make This Possible

You don't need to write code to set this up. Several automation platforms let you connect your AI tools (like ChatGPT, Claude, or Gemini) to your other business apps — and build workflows visually:

→ See detailed reviews of all three in the AI Tools Directory (available to clients)

Note on data: When you connect tools via these platforms, your data passes through their servers as well as the AI provider's — check their privacy policies if this matters for your business.

Connecting to ChatGPT, Claude, or Gemini directly

If you want to use a specific AI model in your automations, you'll need an API key — a code that lets automation tools talk directly to the AI service (never share these — they should always be kept private).

Here's what that involves:

• Create a developer account with the AI provider (OpenAI, Anthropic, or Google)
• Add payment details — API usage is pay-as-you-go and separate from any ChatGPT or Claude subscription you might have (set spending limits to avoid surprise bills on busy workflows)
• Copy your API key into Zapier, Make, or n8n

It sounds technical, but each platform has step-by-step guides that walk you through it. Once set up, you can choose exactly which AI model powers each part of your workflow.

The AI Layer

What makes modern automation different from traditional "if this, then that" rules is the AI element. Instead of just moving data from one place to another, AI can interpret, summarise, categorise, and generate content as part of the workflow.

That's what turns basic automation into something more useful.

Setting Guardrails

When AI is making decisions or generating content as part of an automated workflow, you need guardrails — rules and checkpoints that keep things from going wrong.

Examples of guardrails:

• Requiring human approval before AI-generated emails are sent to customers
• Setting spending limits on any automated purchasing or advertising
• Flagging unusual patterns for manual review (e.g., an order that's 10x larger than normal)
• Keeping AI out of high-stakes decisions entirely — it can prepare the information, but a person decides

The more autonomous you make a workflow, the more important these guardrails become. Remember: AI can still hallucinate or produce bad responses even inside an automation — the risk doesn't disappear just because it's part of a workflow.

A good rule: Don't automate anything you wouldn't want to explain to a customer if it went wrong.

Getting Started With Automation

Don't overcomplicate it. Pick something simple:

• Automatically logging new enquiries from your website to a spreadsheet or CRM
• Sending a follow-up email when a proposal hasn't been responded to in three days
• Creating calendar events from booking confirmation emails
• Posting new blog content to your social media channels

All of these tools have YouTube tutorials and excellent help documentation. You don't need to figure it out alone.

Keep it simple at first. A two-step automation that works reliably is more valuable than a ten-step workflow that breaks. Build confidence with simple workflows, then add complexity.

Don't over-automate too quickly. Automation should remove friction, not create new problems. Get one workflow working well before building the next.