The Next Evolution of AI: Forget Smarter Models – It’s All About the Data
It’s been a noisy summer in the AI world. Headlines have been filled with doom and gloom: For example, OpenAI’s ChatGPT-5 landing with a thud, and an MIT report claiming 95% of AI pilots are failing. For the sceptics, this is “proof” that AI is just hype.
I don’t buy it. The MIT study looked at just 50 projects, a sample size so small you’d fail a basic stats exam for using it. And as someone who uses AI every single day, I can tell you the benefits are real. Productivity gains. Better insights. Faster delivery. It works if you know where to apply it.
Why GPT-5 Felt Underwhelming
The lukewarm reaction to GPT-5 comes down to two things. First, OpenAI’s messaging missed the mark. People expected a bigger model. Instead, GPT-5 is a ‘prompt router’ which analyses your request and chooses the best model to answer. Early glitches aside, results are improving quickly. Mainly because users had never tried a reasoning model before. In fact, pre-GPT-5, just 7% had touched the reasoning models. Now it’s 23%. That’s a huge shift, and users are already seeing better outcomes.
Second, I feel we’ve hit the law of diminishing returns. These models have already chewed through almost every scrap of public data. Gains are now incremental, going from 90% accuracy to 95%. Useful, sure, but compared to the jaw-dropping leap we saw with the first reasoning models, it feels…well, underwhelming.
The Next Battleground: Data
Here’s what I think will be the next big focus for AI for the rest of this year and into 2026. Evolutions in AI won’t come from “smarter brains.” It’ll come from better diets. Whoever feeds their models the most valuable, unique, context-rich data will win.
The major LLM builders – including OpenAI, Anthropic, Meta, and xAI – are expected to compete for exclusive data partnerships, as that’s where differentiation lies. And the same applies to your business. Want your AI to actually add value? It needs to understand your world: your language, your processes, your customers. Microsoft gets it. September’s Copilot update lets you build custom dictionaries with your company’s terminology. That’s just the start.
However…Can You Trust Your Data Is Safe?
Here’s the kicker: if your data is the differentiator, then protecting it is essential. Lose control of it, and you lose your edge. Worse, you gift it to competitors who trained smarter and secured better. That’s why security isn’t just an AI afterthought; it’s the heart of your AI strategy.
At Teneo, we’ve built a framework that does exactly this: securing access to AI, protecting AI workflows, and hardening the infrastructure beneath it. Because if you’re serious about AI, you need to be even more serious about safeguarding the data that powers it.
So ask yourself: are you giving AI enough to work with, and are you protecting that edge from everyone else? If you’re not sure, maybe it’s time we talked. Schedule a consultation with Teneo and see how we can protect your data while unlocking the value of AI.