Human-level AI is here - What's next?

AI reaching human-level intelligence isn’t just another tech milestone—it’s a game-changer for how we understand machines and ourselves.

For years, artificial intelligence has been advancing in narrow, task-specific ways, like playing chess, creating copy and creating videos.

But now, with systems like OpenAI's o3 achieving human-level performance on a general intelligence test, we’re witnessing a shift toward machines that can think more like us—adapting to new challenges, reasoning, and even learning from limited examples. This breakthrough doesn’t mean we’ve quite cracked the code for Artificial General Intelligence (AGI), but it’s a signal that we’re getting closer.

So, what does this leap mean, and why should you care?

Let’s break it down.

Understanding General Intelligence in AI

Artificial General Intelligence (AGI) refers to a machine's ability to understand, learn, and apply knowledge across a wide range of tasks—essentially exhibiting human-like cognitive abilities.

Unlike narrow AI, designed for specific tasks (like playing chess or recommending movies), AGI aims to handle any intellectual task that a human can. I’ve written about this before - there’s a refresher for you here.

The ARC-AGI Benchmark

You can’t measure intelligence without a benchmark, and for AI, the ARC-AGI benchmark test is like the human IQ test.

It appears simple but is challenging: predict what follows in a grid pattern.

Humans, even kids, can swiftly identify patterns such as "adding blue squares in corners" or "forming a border around shapes," However, AI systems have struggled with this task, with the most advanced models achieving only about 5% accuracy before now. This challenge requires machines to think creatively, similar to how humans approach new and unfamiliar problems.

OpenAI's o3 Model: A Significant Leap

The benchmark test sets a high bar, and OpenAI’s o3 model clearing it is a significant milestone.

It scored 85%, which aligns with average human performance and surpasses previous AI models, which scored around 55%.

This means the model isn’t just mimicking intelligence—it’s demonstrating a capacity to learn and adapt in ways that feel genuinely human.

Achieving human-level performance in AI is not all sunshine and rainbows—it comes with many challenges and responsibilities. (Side note: I read a children’s story about a post AGI world over the holidays that was pretty optimistic!)

When you think about how quickly we’ve hit this point, I have already started to wonder what my first post of 2026 and beyond might look like.

If you can remember back to the beginning of the Internet, to the dawn of social media, there were huge amounts of development and huge amounts of excitement, and nobody could really predict how they would impact years down the track. I’m talking about impacting the outcomes of elections, how the algorithm would shape our behaviours, the personal privacy invasions. That a term like “brain-rot” would be named the Oxford Word of the Year as a result of these platforms. The list goes on.

When someone like Geoffrey Hinton, who is widely considered the godfather of AI, says he doesn’t regret his contributions to AI but wishes he had considered safety implications earlier and is now advocating for a 30-fold increase in investment for AI safety research.

So, while everyone will likely be managing a team of AI agents by the end of this year, it’s still up to us humans to reason between right and wrong, handle the messy, unpredictable scenarios outside controlled tests and ensure that human values are aligned.

So, what do I do with all of this?

Let’s get practical for a moment with a few items for your start of year AI checklist:

Stay informed on AI developments
Keep a regular schedule for reading about new AI models and milestones. This will help you identify emerging tools relevant to your work. You’re already here, so that’s a big tick. If you’re not part of the community yet, jump in here. If you have a friend or colleague, please share this email and get them to sign up.

Evaluate potential tools
Compare models and tools for tasks such as market analysis or lead generation. Monitor the results and decide if they offer tangible benefits.

Implement safety checks
If you use AI-driven platforms, explore methods to maintain data security and client privacy. This may include reviewing the provider’s guidelines or setting up your own office safeguards and or policies.

Offer informed guidance to clients
Clients might ask about how AI can affect their property decisions or how you are using AI. Learn the basics of AI’s impact, so you can clearly and confidently address questions from them.

Always combine AI with human oversight
Ensure that important decisions still involve your professional judgment. AI can provide suggestions, but the final call should remain your responsibility. For example, don’t let an AI make employment or leasing decisions without a clearly documented process that has a human in the loop.

Be human when you have the opportunity
I still am not a believer in letting an AI call your clients. Your clients want to hear from you. They want to know you have ‘got them’ in every real estate situation. Pilots have used AI for years to fly planes, but if you’re flying through turbule

nce, you want to hear from the human, not the AI. Like many things in life, just because you can - doesn’t mean you should.

Plan for future skill requirements
Keep an eye on how your role may evolve as AI tools grow more capable. This could mean upskilling or exploring new ways to deliver value in a changing environment. At a minimum, take my free AI crash course.

So, is this a new age of reason?

Going back to yesterday’s post, I felt compelled to ask o3 that question and here is the answer:

We’ll get back to prompting tomorrow - on that note, I have a cracker for a start-of-year planning prompt.

Happy hunting 🚀