Looking Ahead: AI in 2025 and Beyond
The Information asked readers how they envision the progress of artificial intelligence in 2025 and beyond. How much do we already know about AI’s potential and what remains unknown? What are the implementation lessons coming from 2024, and are companies using them to make changes going forward? Is AI going to pay off investors, corporations and employees?
The 249 readers of The Information who participated in the survey (see the methodology) stressed the speed of the AI-generated transformation and the creative destruction it’s causing. They noted that amid this creative destruction it is key to stay level-headed, focus on a business case before technology and work toward measurable returns on AI investments.
Unstoppable: AI Is Hot, and Its Future Is Now
AI will continue to reign as the technology of the day through 2025, believe most of The Information’s survey respondents. Just 10% think another hot technology will replace AI next year.
AI-related change is happening fast, and it’s unstoppable: The biggest group of readers, 43%, believe we already know the impact of AI, while another 44% think we will know by 2030. “The adoption of AI will be Schumpeterian, violent and clear,” writes one reader. Agrees another: “The rate of change from AI will be faster than any previous technology. What took the internet 30 years to achieve will take AI five to 10 years. That’s both exciting and terrifying.”
Confronted by the speed of the creative destruction driven by AI-related innovation, many readers commented on the importance of keeping control. Some stressed the need for regulations, even though the survey did not explicitly ask about regulating AI. “Global regulations must be established by having each country publish policies on what AI is and what it’s allowed or not allowed to do,” writes one reader.
Another reader chimed in with a big-picture framework for securing the future of AI: “The key to the future of AI lies in ensuring that it’s developed with human agency as a central tenet. It is up to us. We must embrace a position of intellectual humility regarding both the prospects and potential of AI.”
Jonathan Bikoff, head of partnerships at Personal AI, an AI startup building digital twins for subject matter experts, compares AI to an engine—a technology that one can build other innovations on. “When the internal combustion engine was created, it wasn’t common knowledge right from the start that it will be used in cars, trucks or airplanes,” says Bikoff. “When the internet was created, people knew it was going to help spread information across the world quickly and securely. But did we know that email, search, the App Store or the iPhone was all going to be built on the internet?”
Similarly, we are in the very early days of AI, also a foundational technology. What we know now, points out Bikoff, is that AI is very good at information retrieval, summarization, synthesis and analysis. What next? More will be revealed about the potential of AI with hyperscaling, training AI on bigger clusters, and refocusing AI from latency to deep thinking, says Bikoff.
Enterprise AI: Lessons Learned Versus Lessons Implemented
The No. 1 lesson about AI implementations learned from 2024 is to start with business, not technology. That means finding out what’s broken that AI can fix better than how it’s handled currently or finding new things to do with AI. “People have been force-feeding AI to date, many times without a real business case,” writes one reader. “AI will not become a powerful, successful tool until real business solutions are found and implemented, with data and outcomes that can be trusted.”
“There are already too many AI products and implementations. This will continue to split the market, confuse buyers and delay implementations,” agrees another reader. But new AI uses will continue to proliferate. According to The Information’s survey, 72% of readers believe that the number of use cases for AI will increase or increase significantly in 2025.
To achieve business value with AI applications, it’s key to implement AI solutions where they can solve existing friction in the work processes, and to monitor if they are adopted and how well they are performing, says Christophe Martel, CEO and founder of Fount, a company dedicated to the examination of daily work and improving productivity. The biggest question for chief technology officers, says Martel, is “How do I figure out where there's a combination of great business need and a great amount of friction in the current way of doing things that AI could alleviate?”
Fount’s software as a service platform measures the user experience at work, helping diagnose potential use cases for AI applications. The company’s solution creates a data-driven model to describe the friction people experience in their work processes. Fed by micro surveys, Fount’s product helps untangle friction in workflows caused by a combination of systems, tools, processes, organizational structures and people. Most of its customers use the platform to figure out how to accelerate AI adoption. It can help designate the areas where AI would be the biggest productivity booster and create a subsequent feedback loop by monitoring how well AI solutions are adopted and how they contribute to business value.
Having a business case first is not the top lesson companies are widely implementing, with just 39% of readers saying they are doing so. A reader with experience in computer vision and AI operations writes that some implementations of AI are difficult, with a high barrier of entry, while others are implemented rapidly and with ease. “The commercial value that AI can provide needs to be weighed individually for each use case,” adds the reader. “People have to determine: “Is the juice worth the squeeze for what I’m trying to do here?”
The second most important lesson learned is that the success of AI starts and ends with the quality of data (18%). It is the No. 1 lesson enterprises are implementing (59%). However, while they are implementing solutions to improve data management, just 17% of readers believe many of the issues around AI-related data quality will be resolved in 2025.
The issue of data quality is also tied to trust. The lesson that AI cannot succeed without trust in AI-generated outcomes comes in as the third most important in The Information’s survey. Efforts to build trust in AI are the second most widely implemented lesson at 51%. But that trust will not be built overnight. Just 24% of The Information’s readers who responded to the survey think in 2025 people and organizations will learn to trust AI-generated outcomes.
One of The Information’s readers echoes the opinion about the importance of trust: “AI only has a future if the data used is trustworthy and the use of AI is transparent to the customer. The degree of misinformation, deepfakes and scams will only accelerate while AI is adopted without care.”
The Payoff: Is AI a Good Investment?
Venture capital will continue pouring into AI startups, with 63% of The Information’s readers saying this inflow will rise or rise significantly in 2025. About half of the readers think the stock market value of AI-driven equities and mutual funds, as well as the price to acquire an AI startup, will also rise in 2025.
One of The Information’s readers is concerned about the potential of an AI-related market crash. “AI will drive a number of breakthroughs. At the same time, AI investment has all the hallmarks of a bubble. By the calendar of the dot-com bubble, we are probably in 1996 or 1997,” writes the reader. “We are a few years from a reckoning where companies making huge investments will have to demonstrate results.”
Another reader of The Information, with experience in providing financial solutions to technology companies, believes there will be massive AI startup consolidations. “AI startups are frequently pitching me, and it’s becoming increasingly apparent that many are just vaporware and are lacking real proprietary IP that is differentiated,” notes the reader.
Ravin Thambapillai, co-founder and CEO of Credal.AI, which builds secure AI assistants for enterprise operations, draws a differentiation between individual AI companies and the AI sector. “If you're investing in the top 1,000 AI startups, 995 of your investments are probably going to be horrifically overvalued and may well go to zero,” he says. “But on the other side of the equation, you see companies that are highly valued, but their growth is phenomenal. The medium company in the AI industry is overvalued. Is the industry overvalued? Definitely not. If you have the luxury to index the top 1,000 AI companies, then you will have great returns.”
What has also become clear, notes Thambapillai, is that the initial belief that technology incumbents will easily add AI features into their existing solutions and there will be no room for startups to succeed off the back of AI was utterly wrong. “There’s a lot more to building an AI product than the actual AI itself. The foundation models are just one layer of the stack,” says Thambapillai. “The question is how you integrate that foundation model into a product that delights the user. The solution is nuanced, and a lot of it comes down to the data. And startups are very good at iterating quickly to build the best user experience.”
Enterprises will continue to invest in AI, with 74% of The Information’s readers saying such investments will increase or significantly increase in 2025. “AI has all the hallmarks of being what I would consider the ‘next computer.’ The level of investment, and the readiness of people and organizations, is on par with what was being done in the ’60s by companies like IBM,” writes a reader.
At the same time, seeing return on that investment will take time, with just 31% of the readers believing ROI on AI will increase or increase significantly in 2025. A reader who works with big organizations on implementing AI notes it’s not easy for them to transform the fabric of how they work with data. “We won’t see major progress even this year,” says the reader. “Maybe in the next three to five years, though it would be great if it happened faster.”
There are also AI skeptics who are not yet convinced about the technology’s economic potential. “AI has yet to produce a viable use case that hasn’t already been handled in other ways. Current implementations of AI are massively expensive with no clear path to revenue,” writes one reader. Another agrees: “The continuous funding of gen AI relies on liquidity and playing tricks with the metrics that measure the gen AI ROI.”
Still, more than half of The Information’s readers (55%) believe AI will lead to an increase or significant increase in productivity and efficiency in operations and logistics. “AI is the last frontier in the journey of technology interventions geared toward increasing executive productivity that started in the early ’90s,” notes one reader. “AI will trim the middle management layers, turning large enterprise teams structurally similar to startups.”
Conclusion
AI will continue to reign as the technology of the day through 2025, believe most of The Information’s survey respondents. How well AI will realize its potential will be up to whether business, individuals and regulators agree on how best to control and utilize it. What we know so far:
AI will surprise us. While it may be moving faster than other technologies and we may know its potential within five to 10 years, we are still at a foundational level of AI. We have an AI engine to build on, and the process of building is just gearing up with hyperscaling and the focus on critical reasoning.
It will take time to learn lessons from AI implementations. Organizations and business leaders are learning some lessons about AI better than others. They are better at improving on the issues related to data quality and trust, which are both integral to achieving success with AI. But they are still struggling with defining and delivering on a business case for AI.
AI will deliver. Funding will continue to flow into AI even though the specter of the dot-com bubble looms over the industry due to its high valuations. Those who take a long-term view and index their AI investments should benefit handsomely.
Methodology: This report is based on a survey of 249 of The Information’s readers, conducted in November 2024.
Company size: 47% of respondents came from companies with less than $10 million in annual revenues, 16% from companies with revenues from $10 million to $100 million, 9% with revenues from $100 million to $500 million, 5% from companies with revenues between $500 million and $1 billion, and 23% with revenues of $1 billion or more.
Industry: Respondents came from across all major industries, with the biggest groups representing technology, media and telecommunications (36%), professional services (16%), and the financial sector and capital markets (14%).
Functional area: Respondents came from across all functional areas, with the biggest groups from general management (33%), information technology (13%), research and development (11%), and marketing and communications (10%).
Title: Respondents’ titles ranged from CEO to employee, with the biggest groups being CEOs (19%), directors (13%) and CEO-owners (12%).
