As top developers continue to work towards achieving Artificial General Intelligence (AGI), a new realization is dawning on them: building the best, most cost-effective, niche enterprise AIs is where the real money lies. This is especially true for knowledge work like law, accounting, investment, and more, before moving on to embodied AIs for manufacturing and other physical tasks.
But what does it take to win the AI race? According to top AIs, bottlenecks in accuracy and data are expected to be solved within the next two years. This means that accuracy and curated data will soon become standardized, tradable commodities. So, what’s the deciding factor in which AIs perform best at knowledge enterprise tasks?
The answer lies in IQ equivalence, or how well these systems process the data. While more data theoretically means more powerful AI, for the vast majority of enterprise tasks, competing developers will have sufficient data very soon. The real differentiator will be how well these systems can process that data.
Take ChatGPT-5, for example. It proved to be a disappointment for many, perhaps because it focused on integration rather than IQ equivalence. As a result, it only eked out Grok 4 on Humanity’s Last Exam, and underperformed it by a substantial margin on the ARC-AGI benchmark, two metrics highly correlated with IQ equivalence.
The takeaway is that top developers seem to be chasing the glory of AGI, at the expense of the IQ equivalence that will probably determine who wins the 2025-26 AI race, and even who gets to AGI first. It’s time to shift focus to what really matters: building AIs that can process data with superior intelligence.
What do you think? Will IQ equivalence be the key to unlocking the true potential of agentic AI?