Weekly AI News Roundup: What OpenAI's Latest Update Means for Your Business

Introduction

This week, OpenAI dropped something that could actually change how you run your business. There’s also a startup charging just £50,000 for work that normally costs half a million, and teachers in Northern Ireland getting 10 hours of their week back. Let’s break down what matters.

Welcome to AI in Business. I’m James Anderson, and this is your weekly AI news roundup. Every week, I condense the biggest AI stories into what actually matters for your business. No jargon, no hype, just what you need to know.

Let’s get into it.

In a nutshell: OpenAI’s GPT 5.4 brings financial plug-ins that let you talk to spreadsheets. A startup is doing due diligence for £50k instead of £500k. Northern Ireland teachers save 10 hours per week with AI. Zoom’s AI avatars raise authenticity concerns. Google expands its AI tools for Workspace. And NVIDIA enters the AI agent race.

OpenAI Launches GPT 5.4 with Financial Plug-ins

OpenAI just launched GPT 5.4, and this one is definitely worth paying attention to. It comes in two versions: Thinking and Pro. The big deal is the financial plug-ins that connect directly to Excel and Google Sheets. So you can basically have a conversation with your spreadsheets. Ask questions about your numbers, and build formulas just by talking.

This is genuinely useful for a lot of businesses. But here is the thing, and this is where I think business owners need to think more carefully. I’ve been using ChatGPT 5.4 through the Codex interface for coding local tools for my business, and honestly, I find it excellent. Fast, efficient, cheap, gets the job done. But what I throw my company’s financials, my trade secrets, my sensitive client data into it? Not on your nelly.

And I think that is the conversation we need to be having. The tool is brilliant for a lot of things, but whether you trust it with your most sensitive business information is a separate decision that each business owner needs to make based on their own risk profile.

AI Disrupting Due Diligence: £50k vs £500k

The next story is exactly the kind of AI application I love to see. A UK startup called Dilligence Squared is using AI voice agents to do something that is traditionally cost a fortune and taken months. They are rethinking the entire due diligence process for private equity deals.

Think about how PE firms work. They want to buy a company, they need to understand the market, they need to talk to customers, they need deep commercial research, and normally that means hiring McKinsey, BCG or Bain. We’re talking £400,000 to £800,000 for a report. And that is per deal.

This startup is doing something different. AI voice agents conduct customer interviews, polling commercial insights, synthesising market data, and they’re delivering the whole thing for just £50,000. That is a tenth of the traditional cost.

And here is the business angle. It’s not just about saving money, it’s about speed and willingness to engage. Because the price is so much lower, PE firms are now bringing this kind of research in much earlier in the process, before they even have high conviction in a deal. That changes how they evaluate opportunities.

And look, if you are a buy and build group or you run a business that does acquisitions regularly, you do not even need to hire a consultancy for this. You could actually engage a developer and build a bespoke tool that does exactly this. Polling customer insights, synthesising data, generating reports, reviewing P&Ls, cash flows. This is the real opportunity here. AI is not just about the big platforms. It’s about taking these old, expensive, bureaucratic processes and rebuilding them from scratch. This is the disruption that actually matters for business.

Northern Ireland Teachers: Ten Hours Saved Per Week

Next, we have a really practical example of AI in action. Six months, a hundred teachers in Northern Ireland, Gemini embedded in Google Workspace. The result: ten hours saved per teacher every single week.

They are using it for parent letters, lesson planning, Irish language resources, all those administrative tasks that eat into teaching time.

And look, I’m not the biggest fan of Google Gemini. I am not going to lie. I’ve used it and it’s fine for general tasks. But here is the thing. It’s cheap because basically Google owns the infrastructure. And in a use case like a school where data is sensitive, it seems like the right choice.

You do not need the most powerful model out there for most tasks. You need something that works, something that is affordable and handles data responsibly. And that is why I normally recommend an alternative is basically running something offline. You own a little server with your own local private LLM running.

And this is what is great about it. Practical use case, automation at its finest. AI being used effectively for repetitive administrative tasks. The teachers should not be spending their time on anyway. Ten hours a week back to focus on actual teaching. That is the kind of AI implementation that makes sense.

Zoom’s AI Avatars: A Step Too Far?

Next up, we have Zoom. And they’ve added some interesting new features to their latest updates. I have to be honest with you about Zoom here. It’s kind of a tool that is standardised for most people. But this latest release seems a little bit weird to me as a business owner.

So they’ve announced that AI avatars can now actually represent you in meetings. These avatars mimic your appearance, expressions, lip and eye movements. I mean, it’s impressive technology technically. But I have a fundamental issue with that approach.

AI is being bolted onto an existing platform. And I get why Zoom is doing it. They are an incumbent trying to stay relevant. But we are moving into a disruptive era where companies do not have to rely on these big SaaS platforms to add AI features for them.

For us, when we record videos, we use tools like Google Meet. And Google Meet comes free with our Google workspace platform. So it’s fine. But I know that large companies will pay extra for Zoom.

And what we do is record all our meetings on Google Meet. We’ll then run these through a local offline AI model that we use on one of our servers. We then extract all the information, transcribe it, feed it into our own note system and our dashboard and calendars for follow ups. And we build our own workflow because we can.

First of all, it means it’s a lot more secure. It means it’s custom to our business and what we actually need. And third of all, it’s basically completely free. Also, one of the advantages is we’re not sitting around waiting for Microsoft, Google or Zoom to add AI features that we actually need. You can build your own tools now, or at least you can hire someone to build them for you.

Also, just going back to the Zoom point, I’m not too sure that I’m super happy about my staff actually being represented by avatars on Zoom meetings. I wouldn’t be too comfortable introducing an avatar to one of my main clients. Would you honestly trust an AI avatar to represent you in your business meeting? Not for me.

Google’s New “Help Me Create” Features

Now, some quick hits. Google has also just rolled out “Help me create” across Docs, Sheets and Slides. It uses your Gmail account and your Google Drive to draft entire documents. The wider review says that it’s great at corporate speak, but a bit bland.

And look, I have to be honest, despite Google basically owning the internet, their AI tools are honestly not the strongest in my view. There are some great tools like I use Google Stitch, which is really good for sort of flash and websites. But I do understand that if you are already paying for Google Suite, this is all essentially free to add on. And it’s not a bad place if you want to dip your toes into AI assistance and just understand at a really basic level. But just remember, obviously with Google Suite and any of these workspaces, the prices do add up. And if you’ve got more staff, then it’s more accounts and a bigger subscription.

NVIDIA Enters the AI Agent Race

Now, this one is worth paying attention to. NVIDIA, the graphics card company, is launching its own AI agent platform.

So here is why this matters. NVIDIA is one of the biggest names in hardware. If you don’t know, I will be covering NVIDIA in more detail in a future article. I’ve also got some stocks and shares with them. And it’s done really, really well over the past two years because of the absolute explosion of the hardware that they’re selling throughout the AI industry.

Anyway, back to the article. So predominantly, they focus on hardware. They also do some software solutions with their CUDA software and drivers for running the GPUs. But what makes this really interesting is they’re building an AI agent platform. And they’re going to be pitching it to enterprise companies like Salesforce, Cisco and Adobe.

The platform will be open source, which means anyone can use it. We like open source software because it gives us the chance to basically fork the code, have a play about with it, customise it and fit it for our business, which is great. That is exactly the kind of approach we advocate.

But because it’s open source, it means anyone can use it. And it’s going to be designed for things like email triage, HR queries and running autonomous helpers in the background. They’re also building in security and privacy tools, which is important because AI agents running wild on your work computer is a real concern.

What I like about NVIDIA is obviously the security aspect. I mean, these guys are building some of the best hardware in the world. And I think because they’re dipping their toe into software, they need to make sure that it’s super, super watertight. Even as an open source model. So I’m really looking forward to trying that.

An open source alternative to OpenAI’s agent platform from a company with NVIDIA’s resources hits the market. It could be a real big deal for businesses that want to deploy AI agents without relying on big tech platforms.

Key Takeaways

• OpenAI GPT 5.4’s financial plug-ins are genuinely useful for businesses, but consider your data privacy posture before uploading sensitive financial information to any cloud-based AI tool.

• AI is disrupting expensive professional services. What used to cost £500,000 in due diligence can now be done for £50,000. This changes the economics for businesses doing acquisitions.

• You do not need the most powerful AI model for every task. The Northern Ireland teachers example shows that affordable, practical tools can deliver massive time savings for repetitive administrative work.

• Building your own AI workflow can be more secure, custom and cheaper than relying on big SaaS platforms to add AI features over time.

• NVIDIA entering the AI agent space is significant. An open source alternative from a hardware giant could give businesses more options for deploying AI without depending on big tech.

Over to You

So the distilled takeaway for this week: if you feel like there’s a lot of AI options in the market already, buckle up because there are even more coming. NVIDIA now joining the scene, OpenAI buying OpenAI, and then you’ve got Facebook, Meta, who just bought Moldbot. Yeah, interesting name.

This is really an open commonspace. I will be covering OpenAI in more depth soon, cutting through the hype and giving you practical examples of how you can integrate it into your business.

And although again, another week’s passed by and AI still seems to be moving at a breakneck speed, the practical advice still remains the same. I would look at picking one repetitive task in your business, test a free AI tool on it, and then track if you can save some time. That is really the basic foundation of everything we cover here.

What AI task in your business would you most like to automate? Drop a comment below.