Budget Finance Assistant in Australia

Below is the script for above video:

Hey everyone, welcome back to the channel! Today, we’re diving into something super exciting: how you can connect a large language model—think of it like an AI brain, similar to ChatGPT or Grok—to your financial data. We’re talking about unlocking powerful insights with just a few simple prompts, and I’m going to walk you through it step by step in a way that’s easy to follow. So, grab a coffee, and let’s get started!

First, let me talk about a trending name: MCP, or Model Context Protocol. In just a few minutes, I’ll break it down in plain English. First off, we all know AI is very powerful—it can write poems, explain history, and even search the web. But when you ask it to send an email from your address or analyze your financial data, it’s stuck. That’s where tools come in. Developers have been hooking up AIs to tools like APIs to make them more useful—like giving them the ability to search the internet or add data to a spreadsheet. But here’s the problem: every tool speaks its own “language,” and connecting them to an AI is like trying to glue together a bunch of mismatched puzzle pieces. It’s messy, time-consuming, and if one tool updates, the whole system can break.

This is where MCP comes in, and it’s a game-changer. MCP is like a universal translator for AIs, created by Anthropic. It’s a standard that sits between the AI and all these tools, making it super easy for them to talk to each other. Imagine if every tool spoke the same language—suddenly, connecting them to an AI becomes smooth and seamless. For example, with MCP, you could tell an AI, “Add an entry to my database,” and it just works, no fuss. MCP is like a layer that translates all the different “languages” of tools into one that the AI understands.

Why does this matter? MCP means now you can connect your own system, like CRM, accounting software, or website, to any AI app with an MCP client, like Claude Desktop. This makes your AI more powerful and flexible. Businesses can build apps that integrate with tons of tools effortlessly, creating smoother, more capable AI assistants.

In short, MCP is a standard that makes AIs way more useful by letting them connect to tools easily, like USB for AI. It’s not some crazy new tech—it’s just a smarter way to make AIs work better for us.

Picture this: you’ve got an app called Claude. It’s awesome because it supports MCP. This MCP interface lets Claude connect directly to your financial tools, like Xero, which is what we’re focusing on today. I’ve already set up the connection—don’t worry, I’ll make a separate video to show you how to do that—but for now, let’s see what happens when this AI brain meets your Xero data.

So, here’s the setup. At the bottom of the Claude app, you’ll see the MCP server I’ve connected to. It’s linked to Xero, and next to it, there’s a number: 42. That’s the number of functions this Xero MCP server can handle—things like creating invoices, listing accounts, tracking payments, or even generating profit and loss reports. Basically, the AI knows about these 42 functions and can pick the right one based on what you ask it to do. Pretty smart, right?

Let’s try it out with a super simple prompt. I’m going to type: “Please use my last month’s Xero bank transactions to list a summary of the top five expense groups by contact.” That’s it! I’m just telling the AI I want to look at my bank transactions from Xero, and I’m leaving it up to the AI to figure out which of those 42 functions to use and how to process the data. It’s like having a conversation with a super-smart assistant.

Now, when I hit enter, something important happens. The app asks for my approval before accessing Xero. This is a great safety feature because you don’t want AI poking around your systems without your okay. I’ll click “Allow Once” for now, but you can also choose “Allow Always” for the session if you’re running multiple queries. Just a heads-up: “Always” only applies to this specific session, so it’s still secure.

Okay, the AI’s now pulling bank transactions from Xero. You’ll notice it might need to make multiple calls to get all the data because Xero sends it in batches—say, 100 records at a time. Once it has everything, it starts analysing. Since it’s May right now, “last month” means April. The AI knows it’s not great at raw calculations—like, don’t ask it to add 1 + 1 and expect a perfect answer—so it writes some code to crunch the numbers accurately.

And here’s the result! The AI gives me a summary of my top five expense groups by contact for April. For example, I spent a ton with Microsoft, followed by some other vendors like Namecheap and Vultr. It’s all laid out clearly, and it took just one simple prompt to get here. How cool is that?

But let’s take it up a notch. I want to visualize this data, so I’ll type: “Please give me a bar chart.” Boom—the AI jumps into action, generates the code, and in a few seconds, I’ve got a clean bar chart showing my expenses. If I click on the bars, I can see details, like how many transactions make up each category—Namecheap had 16 transactions, for instance. It’s interactive and super easy to use.

Now, let’s push the AI a bit further. I’m going to ask: “Can you compare my expenses between April 2025 and March 2025?” The AI starts pulling March’s transactions, and since this is a comparison, it might take a little longer. No big deal—just enough time to sip that coffee! The AI’s also smart enough to handle currency differences, like converting Australian dollars to US dollars for a fair comparison.

And here we go: the results are in. The AI shows me a bar chart comparing March and April, and wow, there’s a big difference. March had much higher expenses across the board. The AI even throws in some insights, like “Spending decreased” and “Expenses dropped.” If I want to dig deeper, I can keep asking questions, but for now, this is already super helpful.

So, let’s wrap this up. What’s the big takeaway? With a tool like Claude and a connection to Xero via MCP, you can use plain English to ask questions about your finances, and the AI does all the heavy lifting. It picks the right tools, pulls the data, runs the analysis, and even creates visualizations. Think about the time you save! Setting this up took me about two hours, and the cost is low—around 40 Australian dollars a month for the Claude and Xero subscriptions combined. For that, you get a virtual finance assistant that’s ready to answer your questions anytime.

That’s it for today’s demo, folks! If you found this helpful, give the video a thumbs-up, hit that subscribe button, and let me know in the comments what else you want to see. Until next time, keep exploring, and I’ll catch you in the next video. Thanks for watching!

A small business owner utilizing AI tools for data analysis

The journey to success is just beginning, and with the right tools and support, your small business can thrive in the age of AI.