A company older than the light bulb just out-AI'd OpenAI for their industry. Here's why that matters to your business.
Thomson Reuters was founded in 1851. They survived two world wars, the internet, and the death of print. Last month, they did something that should get every business owner's attention: they launched their own frontier-class AI model — trained entirely on their proprietary legal, tax, and financial data. Not a ChatGPT wrapper. Not a plugin. A purpose-built model that knows their domain the way a 30-year veteran attorney knows case law.
That move didn't make as many headlines as the latest OpenAI release. It should have.
What Thomson Reuters Actually Did (And Why It's Different)
Most companies using AI today are doing one of two things: typing prompts into ChatGPT, or bolting a chatbot onto their website and calling it a day. Thomson Reuters did neither.
They took decades of structured, proprietary knowledge — legal precedents, tax codes, financial regulations — and used it to train a model from the ground up. The result is an AI that doesn't need to guess what a tax attorney means when they ask a nuanced compliance question. It knows. Not because it scraped the internet. Because it was trained on the real thing.
That's the difference between a general-purpose AI and a domain-specific model. General-purpose AI is like hiring a brilliant generalist fresh out of college. Useful, fast, capable. But domain-specific AI is like hiring someone who has spent 20 years doing exactly the job you need done — and never forgets a single case, contract, or client file.
General-purpose AI is useful. An AI that knows your industry, your suppliers, your pricing history, your customers? That's an unfair advantage.
Thomson Reuters didn't do this to show off. They did it because they understood that whoever owns the data owns the edge. And that principle scales all the way down to your operation.
The Same Playbook Works for SMBs — Right Now
Here's what I hear most from business owners: "I'm not a developer. I don't have a data science team. This stuff isn't for me yet."
I run an e-commerce and import operation. I've automated large chunks of it myself using AI agents, large language models, and workflow tools like n8n. No PhD. No engineering team. Just a willingness to feed my own business data into the right systems.
The version of the Thomson Reuters playbook that's available to you today looks like this:
- Connect AI to your documents. Your SOPs, your supplier contracts, your return policies — when an AI agent can read and reference those, it stops being generic and starts being yours.
- Feed it your order and customer history. An AI that knows your top 50 SKUs, your seasonal demand patterns, and your most common customer complaints is a completely different tool than one that doesn't.
- Wire it into your existing systems. Your CRM, your inventory platform, your email — when AI sits inside your actual workflows instead of outside them, that's when it stops being a novelty and starts being leverage.
- Build agents, not just prompts. A one-off ChatGPT prompt helps you draft an email. An AI agent connected to your systems can monitor your inventory, flag reorder points, draft supplier emails, and log everything — while you're asleep.
None of this requires building a frontier model. It requires being intentional about your data and deliberate about your workflows. That's it.
The Businesses That Win Won't Be the Ones That Use AI Most
This is the part most people get wrong. The race isn't about who uses AI the most. It's about who feeds AI the right data and builds the right workflows around it.
A competitor who uses ChatGPT 40 times a day on generic tasks is not more competitive than you. A competitor who has connected their customer purchase history to an AI agent that automatically personalizes reorder outreach — that's the one you need to worry about.
Thomson Reuters proved that when you combine a powerful AI with proprietary, structured data, you get something no one else can easily replicate. That moat is available to you too. Your data — your supplier relationships, your pricing history, your customer behavior — is yours. Nobody else has it. The question is whether you're putting it to work.
The businesses sitting on rich operational data and doing nothing with it aren't being cautious. They're leaving a competitive gap wide open for whoever moves first.
Consider what's already sitting in your business right now:
- Years of customer purchase data that could predict your next best promotion
- Supplier communication history that an AI could learn from to draft better negotiation emails
- Product return patterns that reveal quality or listing problems before they become expensive
- Internal documents that could train an AI assistant to onboard new staff in half the time
That's not hypothetical. That's Tuesday morning in a well-automated operation.
What to Do With This Information
Thomson Reuters didn't wait until AI was perfect. They moved when the technology was capable enough to be worth training on their data. That moment has already arrived for SMBs too — the tools exist, the cost is accessible, and the window before your competitors figure this out is still open. Not forever, but right now.
Start by auditing what data you actually have. Then ask what decisions you make repeatedly that could be better informed by that data. That intersection is where your AI strategy lives.
At Maqia, this is exactly what we help real business owners implement — not in theory, but in live operations, with tools that don't require a development team. If you want to see how this works for a business like yours and figure out where to start, book a call with us at maqia.co. We'll map it out together.