We are living through a moment of extraordinary flux. Extreme weather events are upending everything from agriculture to insurance. The global order is shifting, creating new stress points and power centers. Biotechnology is quietly rewriting the rules of life itself.
But against all of that, the greatest disruptions for business next year will come from one thing: artificial intelligence. That was the view put forward by Azeem Azhar in his keynote at the London launch dinner for The WIRED World in 2025, WIRED's annual trends briefing, hosted in partnership with Boston Consulting Group (BCG).
Azhar is the founder of Exponential View, an influential newsletter and podcast that draws on his experience as a researcher, investor, and entrepreneur to demystify the impact of technology on humanity. He foresees generative AI (gen AI), in particular, bringing about radical transformation next year.
“It’s a general-purpose technology, and those only come along every five or six decades,” he said. “I’m talking about breakthroughs like the steam engine or electricity. They fundamentally change the cost basis of an economy, increasing productivity, growth, and prosperity.”
So, as this “cognitive steam engine”—as he terms it—develops in the months ahead, how can business leaders make the best of its potential?
Dive In, Don’t Wait and See
It is now two years since gen AI became the biggest story in tech. Since then, the field has made rapid advances. “This is such a dynamic market,” said Azhar. “There is an ongoing acceleration in the capabilities of these new sets of AI systems.”
On the one hand, the increasing sophistication of gen AI is already helping certain sectors see meaningful returns. On the other hand, it presents a challenge for businesses. What new capabilities will be available in 18 months? What current tools will be rendered redundant? It can be disorienting, making it hard to know when or how to invest. But Azhar counsels against taking a wait-and-see approach—he argues that it’s better to exploit progress as it happens, which requires accelerated corporate innovation cycles.
Paul Michelman, BCG’s editor in chief, who co-hosted the dinner, agreed. “The question is: Do you want to be a passenger, or do you want to be a driver?” he said. When BCG recognized the transformative potential of gen AI, for example, it created a conversational AI agent called GENE to assist with content creation. The company is now considering extending its capabilities to other use cases. “Imagine difficult conversations in an organization,” suggested Michelman, by way of example. “A neutral third party like GENE might be able to listen and offer ways to drive it forward when perhaps the people in the room aren't comfortable doing so.”
Prepare for Two Key Shifts
Azhar sees two crucial developments happening in AI next year. The first is that off-the-shelf gen AI-based tools will become sufficiently mature to go mainstream. “Businesses are going to see products out there that are actually good enough to buy,” said Azhar.
The second development is the rise of AI agents. This is a term that refers to gen AI tools that don’t merely produce content but can take actions with a degree of autonomy to achieve a given goal. A key factor driving this is the recent emergence of “reasoning models.” These can break down complex problems into logical steps, and are especially proficient at handling technical tasks.
Agents have the potential to manage complex, end-to-end processes. Azhar pointed to a recent benchmarking exercise in which two different AI models were given difficult AI R&D engineering tasks, and were found to outperform human engineers at leading AI labs on tasks of up to two hours in duration.
“I spoke to somebody involved in reasoning models at one of the three big frontier labs, and asked them, ‘How long before we get to four hours?’ And they said to me, ‘Well, six months, maybe sooner,’” said Azhar. “In six months, the capability of these models will be at the level of the top engineers in the world.”
Problem Selection Is of Paramount Importance
Businesses that wish to take advantage of gen AI can all too easily overlook the importance of the most fundamental strategic question. “I think sometimes organizations don’t spend enough time thinking about what problem exactly they’re trying to solve and the best way to do it,” said Dorothy Chou, director of the public engagement lab at Google DeepMind, who also spoke at the dinner.
In her experience, fruitful areas are often those where the data is already in good shape. While gen AI model architectures can work with large, unstructured data sets, that doesn’t mean that data quality is unimportant. If you want to ground a foundation model in your organization’s documentation, for instance, a high-quality set of documents means your tool is more likely to produce high-quality results.
Chou also advises being ambitious with the bigger-picture problems that, as an organization, you want to tackle. “As I look at where the money is going in AI, it often goes to the easiest possible, lowest-barrier applications,” she said. “Higher-impact areas like biology, life sciences, health, energy, climate, and education are heavily regulated and actually much harder to break into. So you see a lot of money flooding towards ‘personalizing this’ or ‘personalizing that,’ instead of areas where we actually want to see the greatest impact for society.” Creating societal impact isn’t just good from an ethical standpoint, it can also make commercial sense, helping your organization attract investment and talent, and tapping into new lucrative markets.
Don’t Worry About Scaling Walls
There is talk of gen AI hitting a “scaling wall,” where the current approach of throwing ever more data and compute at large language models to drive performance starts leading to diminishing returns. Although there is heated debate on the topic, Azhar maintained that there is categorically not a scaling wall approaching any time soon.
A more realistic obstacle to progress is simply how quickly very large data centers can be built. “We think that demand will ultimately be met because it will be industrially required,” Azhar said.
Creating Content? Ensure Your Brand Stands Out in a Generic Gen AI World
“It took me 24 hours with ChatGPT to realize, as a content creator, you have no choice but to recognize that this is a transformative force,” said Michelman. Indeed, content creation—particularly in marketing—is one of the current primary gen AI use cases.
But one of the big questions around AI content creation concerns stands out. If brands are using the same foundation models, how do they avoid sounding the same as everyone else? “Originality might be at risk when AI tools become widespread in content creation. Brands could lean towards generic outputs; dependency on AI could standardize tone and style resulting in monotonous messages,” agreed GENE, who had been invited along to the dinner.
The conversational AI agent had some advice for WIRED’s dinner attendees on how to overcome that homogeneity problem. “Harness tech-driven efficiency, while ensuring that human ingenuity and distinctive voices shine through.” In other words, as the tools become more common, it's the human use of the tool—the quality of the prompt, the editing of the outcome—that becomes distinguishing.
So whatever happens with gen AI next year, remember: We’re still in this together.
