10+ Predictions By The Business Engineer For 2025
What are some major counter-intuitive trends that might play out in 2025?
I’ve already explained in detail how this AI paradigm's next development cycle might play out (as a single-directional macro-trend) in the next 10-30 years.
Now, look at 2025 and some counterintuitive trends coming up.
Some of these comprise:
Cognitive Automation.
Reverse Adoption of AI.
Energy Innovation.
Verticalization.
Convergence of AI and AR.
Outcome-based Business Modeling.
AI Coaching.
Infrustructure Paradigm.
Reasoning Race.
And why a tech with a lot of buzz might be a flop (for now).
AI Agents Redefine Automation
In 2025, AI-powered agents are set to be tested across many industries, potentially going from simple workflows to completely autonomous agents.
These intelligent agents will operate as interconnected systems, or "constellations," working collaboratively under human supervision to enhance efficiency and productivity.
Deploying such agents is expected to fundamentally transform organizational structures and operational models, significantly improving service delivery and operational efficiency.
As we’ll see ahead, this might also push for creating new business models.
This era we’re going through might become similar to the Industrial Revolution, where we saw, throughout the 18th and 19th centuries, the emergence of a whole new system based on automating the assembly line.
With a core difference, this will be an "Intelligence Revolution,” where AI, as the “computing layer on top,” will enable a sort of “Cognitive Automation,” which will be imbued initially into a few core industries at lower value-add first, like customer support, and move its way up from there!
Read this for that:
Generative AI Becomes Mainstream (For Enterprise)
Generative AI is poised to transition from experimental applications to core components of business operations across various industries.
An interesting take is while in the tech industry, we’re all quite aware of many alternatives to ChatGPT, as we’ve been playing with the APIs of many foundational AI providers; at a consumer level, ChatGPT leads the pack by a significant margin.
By 2025, as competition will intensify in the Generative AI space, these AI systems will be integral in creating personalized content, driving interactive storytelling, and enhancing user experiences in real time.
The integration will also move toward an enterprise level.
The counterintuitive take here is that, contrary to many other industries, this AI paradigm started with a massive consumer push (with the exposition of AI Chatbots), and only later did we get Enterprise running after it to understand how to adopt the technology.
Usually, a “gradual paradigm” moves from enterprise to consumer in a more linear cycle. Generative AI has come all the way around! Leaving many enterprise businesses unable, confused, and lost during this transition.
This is not a “gradual” but a “breakthrough” paradigm.
If you want to understand how we got here, I’ve explained in detail what things might look like under this new paradigm, starting from the company that was supposed to lead us in the AI era and that instead got (almost) crashed by it!
AI Infrastructure Will Lead To An Energy Innovation Breakthrough
While you’ll hear many articles from established media outlets on how AI is wasting energy (in the short-term, it is), those are very good on the clickbaity side, but they miss the point.
The amount of resources, both in R&D and infrastructure development, the AI Race is spurring and will generate in the coming decade might, on the side of it, propel us toward some technological breakthrough when it comes to energy generation, storage, and distribution.
In short, the “wasted" energy AI is sucking right now might turn into a breakthrough in the coming decade!
We don’t know yet what this will be. Still, in The Rise of AI Data Centers, I’ve explained how major AI companies (hyperscalers) are parallelly investing in key areas, such as Nuclear energy for stable power and Liquid cooling for efficient data centers.
Read also:
EdTech Is Dead, EdAITech Is On!
AI has posed a significant threat to the entire edtech ecosystem.
Global edtech investment fell to $3 billion in 2024, the lowest in a decade, as generative AI tools like ChatGPT and LearnLM disrupted the sector. While companies like Duolingo thrived, others like Chegg and Coursera struggled.
The counterintuitive take here is that while AI has killed for promising the “edtech” industry, it was defined a decade back (the paradigm was “get the expensive education you get from a higher level institution, into a packaged cheap version of it”), there will be the chance for edtech to embrace it, to become something else.
As generative AI exploded, edtech (as it has been defined in the last decade) has turned out to be a transitional sector.
Yet, integrating AI into existing edtech business models will play out, depending on the paradigm each company is running with. For instance, players like Stack Overflow and Chegg have been hit the most by it.
Other players who embrace AI will be able to redefine their core with it, thus experiencing a massive resurrection.
For one thing, AI might make it possible to offer personalized education to billions of people worldwide.
Wasn’t this what the payoff of edtech was supposed to be in the first place?
The Humanoid Cycle Will Be Enterprise-Led
As I’ve explained above, Generative AI, contrary to many other industries, has evolved out of a mass consumer application (AI Chatbots) and is now working backward to enterprise.
Well, humanoid robotics will move in the opposite direction.
As I said, “a breakthrough paradigm” moves from consumer backward to enterprise. While a “gradual paradigm” moves from enterprise to consumer.
Here, “breakthrough” or “gradual” isn’t about the societal impact of the technology (indeed, humanoids might have a much more significant impact on it) but rather the technological progression of that technology.
Thus, many might envision the next humanoid in the home. The progression there will go from enterprise, with assembly line humanoids, working its way back to the consumer.
Verticalization of AI
As soon as ChatGPT came out, it was clear to me that a lot of the value in applying AI to everything was in its “verticalization” or capabilities to have AIs that could go from generalists (like ChatGPT is) to specialized generalists (like an AI Accountant, HR, CS, Support rep and so forth).
Well, that’s where we’re going. The counterintuitive take here is that while foundational companies can offer a horizontal platform to build anything on top of it, it will be quite valuable. My take is among the most valuable companies; there will be the verticalized players.
Or those who can tackle an entire vertical with their AI.
In AI Moats, I explain why this matters:
AI Reasoning Wars
The most interesting aspect of the coming wave of “Reasoning AI” is that while we might have figured out a new scaling paradigm, for now, it is also coming at impressive computational costs.
Indeed, OpenAI’s o3 model, while it demonstrates major advancements in AI scaling with a record 88% on the ARC-AGI test (designed to test more human-like intelligence), also came at extremely high compute costs (the high-scoring o3 comes at $10,000 per computation), making it impractical for everyday use, as of now.
However, advancements in cost-efficient AI inference chips (e.g., Groq, MatX) could help make test-time scaling more economical, improving accessibility for institutions and high-value applications before this would get ready for mainstream adoption.
The competitive landscape of AI development is expected to intensify in 2025, with leading AI developers striving to enhance their models' reasoning capabilities.
Yet, a fascinating, counter-intuitive take is the “AI reasoning” layer, which is becoming embedded in large-scale applications. When it comes to the most advanced level, it might be progressing from enterprise to consumer, as only a few major enterprises will have the financial resources to enable very complex reasoning to solve very complex, valuable tasks or find the answer to very complex quandaries (e.g., you’re a Wall Street firm trying to outsmart everyone else, you might spend millions only to test AI reasoning on a single, complex question! Why not try your shot there? It might be worth a billion if you make it right.)
To understand the AI Reasoning Race, read this:
Rise of AI Data Centers with Specialized GPUs
As we move toward 2025, a critical take is that the GPU is no longer relevant as a single component and architecture.
Why? While the first wave of AI has been built on top of the current GPU architecture, it enabled this first phase of scale to train and serve these AI models well.
Yet, now, not only is a specialized GPU critical (AI GPU), but a whole new architecture for data centers is as important - if not more important - to enable AI models to run on mass consumption/enterprise workloads.
Building a robust and competitive AI supercomputer with NVIDIA’s Blackwell architecture might require 40,000 GPUs, plus all the hardware pieces needed to make them into an AI Supercomputer.
Thus, CapeX needs $3 billion even to enter the hyperschaler space.
Outside that, a key take here is that we’ve passed well over the GPU; right now, not only the AI GPU is critical, but the core, the whole architecture on top, is the real hardware paradigm!
That is why, when NVIDIA launched Blackwell, it wasn’t launching a single GPU but an entire AI Supercomputer!
That is also why when Google is working on its AI Chip, Trillium, the company isn’t focusing on the single chip but on the integration of these into an AI Hypercomputer, which is instrumental to training Gemini 2.0 and used in Google Cloud to support both the integration of Gemini across the Google’s suite of products and to enable other companies to leverage these capabilities.
That represents the core of Google's AI strategy.
Read Also:
Convergence of AR and AI Hardware
It took us a decade (and more) to finally get into the moment when AR might be crossing the chasm of mass adoption. And that is all thanks to the current AI paradigm.
Again, I’ve touched this repeatedly, but AR is one of these industries that, thanks to AI, might finally become commercially viable for the first time after some major failed attempts in the last decade (do you remember Google Glasses?).
Indeed, as we enter 2025, Meta plans to add displays to its Ray-Ban smart glasses by 2025, accelerating its AR strategy.
The updated glasses will show notifications and AI responses, complementing Meta’s Orion AR prototype.
Despite challenges like costs and scalability, Meta aligns its AI and AR efforts to dominate next-gen computing platforms.
AR, as a platform, will be a core part of Meta’s AI Strategy.
And Meta isn’t alone there.
Indeed, the fact that Google might be getting ready with a smart glasses release in 2025 also seems clear from the hiring ramp-up the company has been pushing this year, as it hired more than 100 staffers at AR firm Magic Leap.
The arrangement with Magic Leap will help Google work in parallel on underlying AI assistant, operating system, and hardware device for the next AR race, especially vs. Meta, which is trying to establish itself as the top player there.
If it’s unclear, Project Astra will be the underlying AI assistant powering up future models of Google’s smart glasses.
Of course, Apple is also the giant there if it successfully transitions from the iPhone to a successful AR device.
Let’s keep both our eyes on this.
Read Also why AR will matter in Google’s AI Strategy:
Rise of Outcome-Based Business Models
As AI systems become integral to enterprise operations, there will be a shift towards outcome-based business models.
Companies will move from traditional consumption-based pricing to models that charge based on achieved results, such as tasks completed or processes optimized by AI agents.
This transition will redefine value measurement in business transactions, aligning costs with tangible outcomes and performance metrics.
The counter-intuitive take here is it won’t be as simple as it seems. Outcome-based business models will require a whole consumer ecosystem to develop. That is why enterprise might be a great setting stage for this to happen in the coming years. It will enable the testing of different sets of outcome-based business models depending on very custom commercial use cases. Those who will scale from there to enter the B2B world might be good candidates for reaching consumer-based outcome business models.
Read Also:
Quantum Computing Will Be A Flop (for now)
I’m closing the piece with a pessimistic prediction. With all the excitement around Quantum Computing, as Google unveiled Willow (I was also excited about it), this was presented as a breakthrough:
Yet, I believe this might be a big flop as we go into the next 2-3 years, as the time it might take for us to get something quite interesting, at scale, from Quantum computing will be in the coming decade or two.
Indeed, in the AI Convergence, I’ve explained how, in the next 10-20 years, we might see something super interesting from the encounter of AI and new technologies that never managed to cross the chasm, and Quantum Computing is there, too.
But not for now, not in this decade.
Recap: Some of The Key Takes In This Issue!
Automation Moves Beyond Physical to Cognitive Tasks
AI systems transition from automating assembly lines to automating complex cognitive and organizational workflows.Reverse Adoption Paths
Technologies like generative AI and humanoid robotics exhibit contrasting innovation paths: generative AI shifts from consumer to enterprise, while robotics starts in enterprise and moves toward consumer applications.Energy Innovation as a Byproduct of AI Growth
The intense energy demands of AI spur breakthroughs in energy generation, storage, and efficiency, positioning AI as a catalyst for sustainability innovation.Industry-Specific AI Verticalization
Specialized AI systems tailored for specific industries (e.g., healthcare, finance, supply chain) gain prominence over general-purpose platforms.Convergence of Emerging Technologies
AR, AI, and specialized hardware intersect, enabling transformative consumer and enterprise devices that blend immersive experiences with intelligent interfaces.Cost-Driven Technological Balance
Rising computational costs of advanced AI models drive innovation in efficient hardware and software, emphasizing scalability and affordability.Outcome-Oriented Business Models
Enterprises pivot from usage-based pricing to outcome-based models, redefining value measurement in AI-driven workflows and enterprise ecosystems.The Evolution of Education through AI
Traditional edtech fades as personalized, AI-driven education platforms emerge, promising scalable, tailored learning solutions.New AI Infrastructure Paradigms
AI data centers become critical hubs with specialized GPUs and architectural innovations, forming the backbone of scalable, next-gen AI applications.Competitive AI Race in Reasoning and Specialization
The race to enhance AI reasoning capabilities intensifies, with foundational players investing in specialized reasoning models for enterprise and high-value applications.
With massive ♥️ Gennaro Cuofano, The Business Engineer
