The company expects its artificial intelligence (AI) agents and other new products to generate more sales than its ChatGTP chatbot by that time, The Information reported Wednesday (April 23), citing information OpenAI shared with potential and current investors.

It is not clear what the new products will be, but executives have considered selling ads or charging affiliate fees with which it would get a cut of sales that begin through ChatGPT or its AI agents, according to the report.

OpenAI did not immediately reply to PYMNTS’ request for comment.

The company ended 2024 with revenues of $3.7 billion, which was nearly four times the prior year’s sales, according to the report.

OpenAI expects to spend $46 billion in cash over the next four years and turn cash flow positive in 2029, when it expects to generate about $2 billion in cash, the report said.

The company said April 16 that its latest AI models are poised to bring more capable AI agents to business.

The new o3 and o4-mini reasoning models are the “smartest” it has released to date and represent a “step change” in ChatGPT’s capabilities, according to an OpenAI blog post.

These models represent a big step toward more robust agentic AI systems that can independently execute tasks on behalf of users, the company said.

On April 14, OpenAI Chief Financial Officer Sarah Friar said the company is building an AI agent that can do all the work of software engineers, not just augment their skills.

Friar added that OpenAI is moving beyond being solely a model builder to becoming a comprehensive AI infrastructure provider and applications developer.

“Today, OpenAI is so much more,” Friar said. “We’re going down into data center technology … and we feel like we’re creating a lot of IP there, and it’s really important for us to own that.”

On April 11, OpenAI CEO Sam Altman said the generative AI company has reached 800 million people.

“Something like 10% of the world uses our systems, now a lot,” Altman said.

The post Report: OpenAI Expects New Products to Drive Revenue to $125 Billion in 2029 appeared first on PYMNTS.com.

And many are putting their money on that, with the latest PYMNTS Intelligence data from the April 2025 edition of the CAIO Report, a collaboration with Coupa, revealing that over 80% of U.S. chief financial officers (CFOs) at large enterprises are either utilizing AI in their accounts payable (AP) processes or are considering its adoption.

Specifically, the report found that CFOs are demanding tools that can offer real-time visibility into expenditures, sharpen vendor negotiations and guide more strategic budget optimization. With margins tightening and pressure mounting to do more with less, these three capabilities have become non-negotiable for finance teams operating with innovation.

After all, since its machine learning and data analysis days, enterprise AI has promised to transform how businesses operate by streamlining operations, forecasting with precision and automating decision-making at scale. But for CFOs, the hype has often outpaced reality.

What many artificial intelligence tools lack isn’t power or potential, but purpose. CFOs need tools that solve core problems and give businesses leverage in real time.

If enterprise AI providers can meet three emerging but critical needs: real-time visibility, empowered vendor negotiation and dynamic budget optimization, they won’t just win deals. They could earn trust, shift paradigms and redefine how modern finance operates.

Read the report: Smart Spending: How AI Is Transforming Financial Decision Making

Enterprise AI Essentials for Smart Spending, Smarter Strategy

Legacy ERP systems weren’t built for speed or clarity, but transparency in financial operations is paramount for CFOs aiming to maintain control over corporate spending.

The PYMNTS Intelligence study data indicates that 68% of CFOs are willing to invest in AI solutions that offer real-time insights into expenditures. Such tools enable finance leaders to monitor spending patterns, identify anomalies and make informed decisions promptly.

For example, some AI platforms are now using machine learning to tag and categorize expenses the moment they occur, even flagging unusual patterns like off-contract purchases or sudden vendor price hikes. These systems don’t just report — they interpret. That real-time radar is particularly powerful in industries with complex supply chains or decentralized purchasing.

The result is fewer surprises, faster course corrections, and a culture of financial accountability.

At the same time, effective vendor management is a critical component of financial strategy. AI technologies can assist CFOs in negotiating better terms by analyzing historical data, market trends and supplier performance metrics. By leveraging predictive analytics, CFOs can anticipate price fluctuations, assess supplier reliability and determine optimal procurement strategies.

The study highlights that AI tools are instrumental in managing payment terms and optimizing working capital. These capabilities not only improve vendor relationships but also contribute to the overall financial health of the organization.

What It Means for the Enterprise AI Ecosystem

Static budgets are quickly becoming relics of the past. Today’s CFOs need dynamic, AI-enhanced tools that support agile financial planning — helping teams adjust spending in response to new data, not last year’s assumptions.

The study reveals that 83% of enterprises using AI for accounts payable apply it to at least one payment execution feature, such as payment scheduling or early-payment discount usage.

Furthermore, 74% of enterprises employ AI to support cash flow management, ensuring that funds are allocated efficiently across various departments. By integrating AI into budgeting processes, organizations can respond swiftly to market changes and reallocate resources as needed.

As CFOs grow more tech-savvy — and more skeptical — vendors need to evolve. That means tighter integration with core systems, better explainability of AI models and a focus on the end user. If a tool adds cognitive load or demands heavy IT support, it’s unlikely to last.

Ultimately, CFOs aren’t asking AI to reinvent finance. They’re asking it to make finance work better with smarter decisions, faster insights and stronger results.

The post What Enterprise CFOs Want From GenAI appeared first on PYMNTS.com.

With this new office, Layer 6, which currently operates from its head office in Toronto, will grow to more closely support the bank’s U.S. operations and access an expanded pool of talent, TD Bank Group said in a Tuesday (April 22) press release.

“Our U.S. expansion of Layer 6 underscores our commitment to deepening our presence in New York City and investing in the future of innovation,” TD Bank President and CEO Leo Salom said in the release. “The new Layer 6 office establishes a strong foundation for advancing our GenAI capabilities and bringing critical expertise and delivering in-house.”

When it opens, the new office will house 20 data scientists, applied machine learning scientists, GenAI implementation specialists and others, according to the release.

The bank’s deployment of AI technology focuses on both personalization for customers and streamlined process execution within the business, the release said.

These efforts are driven by Layer 6, which TD Bank Group acquired in 2018, per the release.

Luke Gee, chief analytics and AI officer at TD, said in the release that the expansion of Layer 6 is the next major step in the company’s efforts in banking innovation.

“With recent advances in areas like generative AI, we continue to leverage the potential of this game-changing technology through the work of colleagues across the bank, with Layer 6 continually driving thought leadership and breaking new ground,” Gee said.

Nearly all banking boards have approved generative AI initiatives, according to the PYMNTS Intelligence and NCR Voyix collaboration, “Is AI the Master Key to Banking’s Next Era?”

The report found that 72% of finance leaders actively use AI in their operations, with its applications ranging from fraud detection (64%) to customer onboarding automation (42%).

When TD Bank Group acquired Layer 6 in 2018, the bank described the Toronto-based AI company as “a global thought-leader and pioneer in the delivery of responsive, personalized and insight-driven experiences for the financial services industry.”

The post TD Bank to Open AI Center in New York City appeared first on PYMNTS.com.

Several companies are offering services today to create artificial intelligence (AI) digital twins of deceased loved ones. Think of Superman conversing with the 3D twin of his Kryptonian father Jor-El in the Fortress of Solitude, who continued to give sage advice to his son well after he perished.

Organizations offering AI afterlife services include Re;memory from DeepBrain AI, HereAfter, Character.ai, StoryFile, Project December and even MIT’s Augmented Eternity.

But in a research paper by Google DeepMind and University of Colorado researchers, they put forward a more advanced version: “Generative ghosts.”

Generative ghosts go further than being AI digital twins that think, sound and look like the dead; they can also generate new conversations based on new life events or current news, according to the researchers.

Ghosts in the Machine

Moreover, they will have AI agent capabilities so they can act as you.

For example, they could continue to work after death — potentially removing the need for life insurance since they will keep on supporting dependents in the afterlife.

This is coming, the researchers predicted.

“We anticipate that within our lifetimes it may become common practice for people to create custom AI agents to interact with loved ones and/or the broader world after death,” the researchers wrote.

These generative ghosts are capable of learning, evolving and acting independently in ways that could fundamentally reshape how we grieve, remember and interact with the dead.

People can even create their own generative ghosts as part of their end-of-life planning process, the researchers said.

But consumers are mixed about whether there should be AI afterlives.

According to a 2024 survey by cybersecurity firm Kaspersky, 38% said it was not acceptable to create a digital identity of the deceased, while 35% said the opposite.

However, most were certain that these digital twins would reopen grief: 67% said seeing images or stories about people who have died would be upsetting to those close to them.

Read more: How the World Does Digital: A Global Benchmark of Consumer Digital Transformation

It’s Not Sci-Fi; It’s Estate Planning

The idea is not science fiction — nor is it far-fetched.

Companies offering a more basic form of digital immortality are already around, and their services are becoming more popular, especially in Asia where communicating with dead ancestors is part of the culture.

For example, in South Korea, a video of an emotional mother conversing with a virtual reality recreation of her deceased young daughter drew 36 million views as of this writing.

Startups like HereAfter and Re;memory already let clients record life stories that can later be animated through AI to converse with family after death.

In 2021, Microsoft was granted a patent for a system that would create a chatbot modeled after a person, using their images, social media posts, messages, voice and written content. The chatbot would be trained on this data. But Microsoft’s former general manager of AI programs Tim O’Brien, said in a post on X there were no plans to develop it.

These generative ghosts don’t have to be human; they could be pets too.

Looking ahead, researchers said generative ghosts could inhabit robots so grieving family and friends can physically touch and hold the deceased.

Beyond the social and psychological impact, these AI doppelgangers open the door to legal implications, especially for estate planning. 

Generative ghosts could potentially serve as a new interface for executing living wills, managing digital assets or clarifying posthumous wishes. 

In the future, a generative ghost might be used to weigh in on a contested will and other property and inheritance matters, especially when written documents leave room for doubt.

But this begs the question: Will the generative ghost have the same rights as its human? Are they considered the same being, albeit one is in digital form? Or is the generative ghost a mere representation, with more limited rights?

As a result, we may soon see AI clauses become standard in estate planning documents, much like health care proxies or advance directives are today.

See also: Digital Has Become a Way of Daily Life for 622M Consumers in 11 Global Markets

The Pros and Cons of Generative Ghosts

People might want to create a generative ghost of themselves as a form of digital legacy — a way to share experiences with future generations, guiding descendants through life events, or simply being remembered in an interactive way. Families may seek it for their loved ones — for comfort, closure or help with practical matters like estate administration.

“Our goal is to get rid of the pain of grief,” said Justin Harrison, founder of You, Only Virtual, a company that creates virtual versions of the deceased, in a documentary for German public broadcaster DW that aired this month.

The researchers said generative ghosts hold value for society. They may preserve endangered languages, document lived experiences of historical events or allow students to ‘interact’ with historical figures for education.

But there are major pitfalls too, the researchers warned.

They could affect the mental health of those left behind through delayed grief, or even confusing a simulation with reality.

Security risks include identity theft, ransomware attacks on digital ghosts and even malicious use by the ghost’s creator — just like cybercriminals use malware to infect software.

There are also risks to one’s reputation. What if the ghost hallucinates false memories or reveals secrets the deceased never intended to share? And who controls such representations?

As these AI ghosts become more capable, society must grapple with economic, ethical and spiritual implications.

“Widespread economic activity by generative ghosts might impact wages and employment opportunities for the living,” the paper said.

Also, should religious institutions embrace or reject these digital afterlives? Might new spiritual movements arise to worship them?

“The emergence of ‘AI afterlives’ may reshape society in complex ways beyond our current imagining,” the researchers wrote.

The post Forever Online: ‘Generative Ghosts’ Live in the AI Afterlife appeared first on PYMNTS.com.

Sources familiar with an xAI investor call last week told CNBC Monday (April 21) that Musk was on the call and is seeking to establish a “proper valuation” for his artificial intelligence (AI) startup.

Although Musk, who was a co-founder of AI pioneer OpenAI, did not formally announce a capital funding round for xAI, the sources for the CNBC report think that is coming soon.

The sources also said a $1 billion potential run rate figure was mentioned on the call.

CNBC said the investor call took place just months after reports that xAI was seeking to secure up to $6 billion at a $50 billion valuation, with the funding reportedly coming from Middle Eastern sovereign wealth funds and backing from global investors. That capital was earmarked for acquiring 100,000 Nvidia chips, CNBC said, a key component in the company’s AI infrastructure.

xAI recently merged with the Musk-owned social media platform X in an all-stock deal. PYMNTS reported that he valued xAI at $80 billion and X at $33 billion.

Founded in July 2023, xAI launched the chatbot Grok last year to compete with other leading AI chatbots, including Anthropic’s Claude and OpenAI’s ChatGPT. The CNBC report said that at the time, xAI touted that Grok’s real-time internet knowledge enabled it to launch with just two months of training.

PYMNTS reported in March that xAI and Nvidia joined a $30 billion AI Infrastructure Partnership  backed by BlackRock, Microsoft, and Abu Dhabi AI investment group MGX, with the ultimate goal to raise up to $100 billion for AI development. The partnership aims to build the next generation of AI-ready data centers and support the escalating demands of generative AI models.

The CNBC report also noted that Tesla, where Musk is CEO, experienced a 40% decline in shares this year and could raise questions about his management bandwidth.

The post Report: New Valuation Push for Elon Musk’s xAI appeared first on PYMNTS.com.

Enterprise AI is different. It is the use of artificial intelligence (AI) at scale within companies to boost productivity, enhance decision-making, empower customer service and level up risk management, cybersecurity and innovation.

Consumer AI has no organization-wide goals; enterprise AI does. Unlike consumer AI applications, enterprise AI is designed to solve complex business challenges at scale and integrate with existing corporate systems.

At its core, enterprise AI refers to AI systems deployed within organizational settings to analyze vast amounts of business data, automate routine tasks and provide actionable insights for strategic decision-making.

“Enterprise AI includes policies, strategies, infrastructure and technologies for widespread AI use within a large organization” that requires “significant investment and effort,” according to AWS.

Examples of enterprise AI include:

• A credit card company using AI to detect fraud across billions of transactions
• A retailer using AI to optimize supply chain logistics
• A hospital using AI to analyze medical images for diagnoses

The most effective enterprise AI implementations typically address specific business problems rather than pursuing technology for its own sake.

Consumer AI can be free or low-cost: think of the free or $20-a-month versions of ChatGPT, Gemini, Anthropic’s Claude, Perplexity and others. Its capabilities can be enough for the needs of small businesses.

Enterprise AI is far more costly, usually charged on a per user, per month basis. This is artificial intelligence tailored for a particular company’s needs and goals.

Typically, enterprise AI is deployed by larger companies at scale and integrated into existing systems. It is hyper-focused on security and compliance, such as GPDR and HIPAA regulations.

Implementation of enterprise AI is typically done in stages, starting with identifying use cases with the highest impact, building proofs of concept and the deploying it at scale within the company.

While most companies already have some form of automation in their existing systems, enterprise AI can make them smarter and more intuitive. Think of using a phone tree (press 1 for account information, press 2 for customer service, for example) compared with a conversation with ChatGPT.

Read more: How to Manage Risks When Employees Use AI Secretly for Work

Tips for Deploying Enterprise AI

To deploy enterprise AI, John Roese, Dell’s chief technology officer and chief AI officer, said in a recent YouTube video that the company first asks itself, “what problem are we solving? What impact do we expect?” Then, “once we figure out what we’re trying to do, how are we going to do it?”

Roese said companies should start by asking themselves these questions:

What makes your company special? If improved with AI, what will actually let you win?

At Dell, Roese zeroed in on three strengths: Secure supply chain, enterprise salesforce and global services capabilities.

What process in each area of your company’s strengths will move the needle if enhanced with AI?

Their enterprise sales people would spend 40% of their time doing research as they prepare to meet clients.

What tools and technologies will you use to solve the issue?

To help its enterprise salesforce, Dell used AI to speed up the preparation time for sales people before they meet with clients.

To prepare for AI deployment, Roese said companies must do the following:

• Organize and structure the data to make it useful for AI.
• Don’t create a custom AI tool for each use case — you’ll end up with a lot — but rather see how the tool can be used for multiple purposes.

Roese said Dell figured out that it only needed five types of AI capabilities that can be broadly applied to all use cases across the company.

The next question is whether to build or buy the AI tool. A year ago, Dell built them. Today, it doesn’t. “There are now sufficient off-the-shelf tools that the vast majority of the AI components are inclusive of the models, the developer frameworks, the engines, [and] can actually be consumed as a standardized piece of technology from a provider,” Roese said.

The post AI Explained: What Businesses Need to Know for Enterprise AI Adoption appeared first on PYMNTS.com.

Last week, the Federal Trade Commission began its anticompetition trial against Facebook parent Meta for its social media dominance. It is seeking a divestiture of Instagram and WhatsApp.

In the same week, a Virginia judge ruled against Google, saying it acted illegally to maintain its dominance in online advertising technology. Google is also the subject of another antitrust lawsuit from the government for its search dominance. The government wants Google to give up its Chrome web browser, the most popular in the world.

“The ongoing antitrust trial against Meta and Google’s recent antitrust losses can substantially reshape the AI ecosystem due to their roles as major AI developers and innovators,” Ron Westfall, research director of communication networks at The Futurum Group, told PYMNTS.

There will be short-term turbulence, but the companies would pivot. It could also lead to “more streamlined and focused innovation,” similar to how AT&T’s breakup led to more telecom and internet innovation and gave rise to competitors that thrived, Westfall said. In 1984, the U.S. government broke up AT&T, widely known as “Ma Bell,” into seven regional telecom companies called “Baby Bells,” one of which is today’s Verizon.

Dev Nag, CEO of support automation firm QueryPal, believes a breakup will lead to more opportunities for smaller companies.

“When monopolistic firms like Google and Meta face breakups or restrictions, we often see an explosion of innovation from smaller players who finally have room to compete,” Nag told PYMNTS.

“The forced opening of critical resources — like Google potentially sharing search data with competitors — could democratize AI development in ways that accelerate progress beyond what any single company could achieve.”

These antitrust actions will likely result in a more resilient AI landscape and create pathways for the next crop of AI leaders who might otherwise be “smothered by the giants,” Nag added.

The AI ecosystem is dominated by a handful of large firms that control critical resources — massive datasets, computing power and top-tier AI talent — according to a study from the Center for Security and Emerging Technology at Georgetown University. These companies also invest heavily in research and development, which fuels advances in everything from consumer products to defense applications.

Apple is also facing its own antitrust trial over allegedly anticompetitive App Store policies, and Amazon is being sued by the FTC for allegedly using its dominant position in eCommerce to stifle rivals.

Read more: Meta’s Landmark Antitrust Trial Opens With Focus on 2020 Election

Breakup’s Impact on Innovation

Meta plays the spoiler in the AI race by offering open-source or freely available advanced AI models that can go head-to-head with those offered by AI leaders like OpenAI, Google, Anthropic and others. Meta’s 2-year-old Llama family of models is the most popular open-source AI models in the world, hitting 1 billion downloads as of March 18.

But Big Tech could become more cautious about investing in R&D under regulatory pressure, which dampens innovation, said Shawn DuBravac, CEO of the Avrio Institute who formerly worked at the Department of Justice’s antitrust division. He pointed to the “significantly reduced” investment by AT&T in its venerable Bell Labs after its breakup. Bell Labs — home to many tech breakthroughs and 11 Nobel Prizes — became a “shadow of its former self,” he told PYMNTS.

Mike Conover, CEO of Brightwave, said that while anticompetitive practices have no place in an efficient market, the deep pockets of Big Tech are needed to keep U.S. AI leadership. “Large-scale language model training benefits from large-scale investment,” he told PYMNTS. “Regulators and courts would do well to preserve our domestic ability to execute Manhattan Project-scale AI programs like those initiated by these companies.”

There is a risk of “regulatory overreach undermining U.S. AI market agility, particularly as China ramps up AI investment,” Westfall agreed, although noting that appeals will likely postpone any breakup action for years.

But Nag thinks otherwise. “Counterintuitively, these antitrust cases might strengthen America’s position against China in AI, not weaken it.” A more vibrant ecosystem with more competitors and resource constraints often produces more breakthrough innovations, he added.

Damian Rollison, director of market insights at SOCi, thinks the U.S. can take a page from Europe.

“The remedies of forced divestiture are blunt instruments that may not achieve the desired ends,” Rollison told PYMNTS. Instead, a better way is how the Europeans regulate, which “protects consumer rights and increases the responsibility of Big Tech over its content and influence.”

The post Breakup of Meta, Google May Spur New AI Innovation Wave, Analysts Say appeared first on PYMNTS.com.

PYMNTS surveyed 60 chief financial officers (CFOs) from United States firms with at least $1 billion in annual revenue, revealing a significant increase in the strategic application and positive outlook for generative artificial intelligence (AI) within their organizations. The findings indicate that CFOs are increasingly leveraging generative AI (GenAI) for high-impact tasks such as data visualization, financial reporting and capital management, signaling a departure from more routine applications. This growing reliance suggests that finance leaders are betting on GenAI to navigate economic uncertainties and enhance their strategic decision-making processes.

The PYMNTS Intelligence research highlights a notable evolution in how CFOs perceive and utilize GenAI. In June 2024, a greater proportion of middle-market firms were employing the technology for medium-impact activities compared to low-impact ones, a trend that began in March. Specifically, 45% of middle-market firms reported using GenAI for medium-impact tasks in June, up from 35% in March. Furthermore, CFOs expressed increasing confidence in the technology’s importance across various financial functions. The percentage of CFOs who viewed GenAI as important for financial reporting surged from 37% in March to 68% in June, and for capital management, it rose from 30% to 58% over the same period. This growing adoption and heightened expectations underscore the transformative potential of GenAI within corporate finance.

Key data points from the PYMNTS Intelligence report include:

• Over 98% of CFOs surveyed anticipate a positive impact of GenAI on their industry within the next three years, particularly in accelerating decision-making, up from the 77% who held the same view in March.
• The most common reported use of GenAI among CFOs was for data visualizations and reports, with over 60% of respondents indicating its application in this area.
• CFOs’ perception of OpenAI as the leading GenAI company has decreased, with 20% identifying it as the leader in June, down from 27% earlier in the year. Meanwhile, perceptions of Microsoft, Google and Meta as dominant players have increased.

The report also explores CFOs’ perspectives on the shifting landscape of GenAI providers. While OpenAI’s ChatGPT remains a prominent name, the movement of key personnel to competitors like Anthropic and the evolving market shares of companies like Microsoft and Google are capturing the attention of finance leaders. This dynamic environment could potentially offer CFOs more diverse and profitable avenues for their GenAI investments as new modalities and use cases emerge.

The post 45% of Middle-Market Firms Report Increased Use of GenAI appeared first on PYMNTS.com.

The startup — whose launch was covered by a Saturday (April 19) TechCrunch report — debuted last week with that mission, drawing some controversy in the process.

“Mechanize will build virtual work environments, benchmarks, and training data to enable the full automation of all work,” Tamay Besiroglu wrote in a post on X.

The company itself took things one step further, promising the full automation of the economy, according to its X announcement.

“We will achieve this by creating simulated environments and evaluations that capture the full scope of what people do at their jobs,” Mechanize said. “This includes using a computer, completing long-horizon tasks that lack clear criteria for success, coordinating with others, and reprioritizing in the face of obstacles and interruptions.”

According to TechCrunch, Besiroglu calculated Mechanize’s total addressable market by aggregating all the wages humans are paid, roughly $60 trillion per year on a global scale.

However, he told the news outlet that “our immediate focus is indeed on white-collar work” and not manual labor jobs that would use robotics. 

The report noted that the response to this launch was “brutal,” with one X user arguing that the “automation of most human labor … will be a huge loss for most humans.” 

As PYMNTS wrote earlier this month, the question of whether AI will make human workers obsolete or make them more valuable is one that concerns many workers.

However, MIT economics professor David Autor argues that AI will in most cases augment workers rather than replace them.

“There are two competing visions of AI. One is machines make us irrelevant. Another is machines make us more useful. I think the latter has a lot to recommend it,” Autor said at the 2025 MIT AI Conference in Cambridge, Massachusetts.

He pointed to a historical precedent. “Over the last 200 years, we have automated so much of what we do. We have mechanized. We have moved ourselves out of agriculture, out of manufacturing, out of back-breaking toil,” Autor said. Yet, “We have made labor more valuable during that period.”

Still, workers are worried. Research by PYMNTS Intelligence — from the report “GenAI: A Generational Look at AI Usage and Attitudes” — shows that 54% of respondents think AI poses a “significant risk” of widespread job displacement. 

These concerns span industries and demographics, though some are more worried than others. Workers in the technology sector and non-customer-facing roles were the most concerned (58%), while healthcare and education workers were less so, at a respective 48% and 52%.

The post AI Startup Mechanize Aims for ‘Automation of All Work’ appeared first on PYMNTS.com.

The new o3 and o4-mini reasoning models are the “smartest” it has released to date and represent a “step change” in ChatGPT’s capabilities, according to an OpenAI blog post. The models make ChatGPT more capable — and more similar to holistic human capabilities.

“There are some models that feel like a qualitative step into the future,” OpenAI President Greg Brockman said in an OpenAI video. He said GPT-4 was one of those models — and now o3 and o4-mini.

These models represent a big step toward more robust agentic artificial intelligence (AI) systems that can independently execute tasks on behalf of users, the company says. With full access to ChatGPT’s tools and custom tools, these models can autonomously coordinate multiple actions to solve complex problems.

OpenAI’s latest models are the latest salvo in an increasingly crowded AI market.

Google DeepMind’s Project Astra, a multimodal AI assistant, comes the closest in capability and was first unveiled a year ago. Astra can see, hear and understand its surroundings. However, Astra is not as advanced in reasoning, is not agentic and hasn’t been released publicly.

In March, OpenAI CPO Kevin Weil said at a conference that while ChatGPT is currently at the top, “it doesn’t mean that we’re going to have a lead forever. I think those days of us having a 12-month lead are probably gone — there’s just too many smart people, too much going on in the ecosystem.”

Whether or not OpenAI stays in the lead, companies are all in on AI.

Nearly 90% of CFOs report that they are seeing a “very positive” ROI from generative AI, according to a February PYMNTS Intelligence CAIO Report. That’s three times as many as those who said so in March 2024.

Moreover, at least 91% of CFOs surveyed have “high” or “complete trust” in generative AI’s output in 10 key areas, partly due to the use of their company’s own data as the basis for the AI’s responses.

However, 29% did say that the AI’s responses “might not be very insightful” — this was the top concern about generative AI’s outputs.

Read more: OpenAI Product Chief Says ChatGPT Will Become Agentic in 2025

How o3 and o4-mini Are Different

Here’s what makes these models different from OpenAI’s other models:

• They can use every tool within ChatGPT, including searching the internet, analyzing uploaded files and other data, reasoning about images and generating images.
• They incorporate images directly into their reasoning, which boosts their problem-solving skills. Images can be blurry, upside down or drawn by hand. The models can zoom into an image, if needed.
• They merge the reasoning capabilities of OpenAI’s o-series AI models with the conversational abilities of the GPT series of large language models (LLMs).
• They reason through which tools to use — one task called for using 600 tools — to solve complex problems, usually within a minute. OpenAI said this translates to “significantly” better performance.

“The reason we’re so excited about tools is that it makes our reasoning models that much more useful and that much smarter,” Mark Chen, OpenAI’s chief research officer, said in the OpenAI video.

One example: A user asks, “How will summer energy usage in California compare to last year?” The model searches the internet for public utility data, writes Python code to build a forecast, generates a graph or image and explains key factors for the prediction, according to OpenAI.

As for performance, the startup said o3 makes 20% fewer major errors than predecessor o1 (o2 was skipped) on difficult, real-world tasks. It especially excelled in programming, consulting and coming up with creative ideas.

The o4-mini model, meanwhile, focuses on balancing performance with efficiency. This smaller model excels in mathematics, coding and visual analysis tasks. Efficiency gains let o4-mini support higher usage volumes than o3-mini, making it ideal for larger and more complex tasks.

OpenAI said for most real-world uses, o3 and o4-mini will be cheaper than o1 and o3-mini while outperforming them on tasks.

The models are now available for ChatGPT Plus, Pro and Team users. ChatGPT Enterprise and Edu users will get them in a week. Free users can try o4-mini by selecting “Think” before entering a prompt. Developers can access the models via the Chat Completions API and Responses API.

OpenAI o3-pro should be out in a few weeks.

The post OpenAI Says Its Latest Reasoning Models Upgrade Agentic Capabilities appeared first on PYMNTS.com.