‘Vibe Coding’ Has Arrived For Businesses And Personal Users

It’s never been easier to create your own app with ‘vibe coding.’ Now, professional software engineers are bringing it into the enterprise. ‘Vibe Coding’ Has Arrived For Businesses And Personal Users

Creating your own app is now possible with any number of artificial intelligence-based tools, leading to the “vibe coding” revolution for code-writing amateurs.

Ultimate Resource On Artificial Intelligence

But professional developers are picking it up now, too, bringing the practice—generally understood as the ability to create functioning apps and websites without strictly editing code—into businesses.

In the next three years, market research and IT consulting firm Gartner predicts that 40% of new software for businesses will be created with techniques involving AI bots translating plain English prompts into usable code.

Combined with the growing popularity of AI-powered assistants and editors, vibe coding’s rise in the enterprise reflects a significant shift in how quickly apps are being conceived and delivered, with many implications for professional developers.

The use of AI to write code has steadily increased among companies since the launch of ChatGPT a few years ago.

For some large tech companies like Google and Microsoft, AI is already writing over 20% of the code instead of human engineers.

Vibe coding can act as a “creative partner” to generate new ideas engineers might not have considered, according to Lauren Wilkinson, Vanguard’s divisional chief information officer for financial adviser services.

Using AI tools like OpenAI’s GPT models and Anthropic’s Claude, Wilkinson’s team is vibe coding new webpages with the help of product and design staff.

The process has eliminated the need for traditional handoffs of work between teams, speeding up the design for a new Vanguard webpage by 40%.

Prototyping went from taking two weeks to 20 minutes, she said.

The team’s roughly 200 engineers are also experimenting with AI-powered code editors like Windsurf and Cursor—a fast-growing software-as-a-service company.

“It’s not just transforming the engineering piece, it’s really transforming the whole process,” Wilkinson said, referring to the practice of vibe coding.

But vibe coding, at least in a professional setting, does not rely on vibes, or intuitive feeling, alone.

AI tools can be prompted with plain English to create a usable foundation for code, but that code isn’t ready for use right away, Wilkinson said.

“The role of the engineer is still very, very critical to make sure that the boundaries and conditions are set up front for what the vibe coding is going to produce,” she said.

“It doesn’t excuse the engineer from needing to understand what’s going on behind the scenes.”

Brian Kirkland, CIO of Choice Hotels, said some of the hotel chain’s employees, including its engineers, recently taught and took classes on vibe coding.

But for vibe coding to really take off, it needs to work inside of a company’s existing code base and connect to its other core development systems.

Once that framework is built, however, vibe coding works as anticipated, Kirkland said. “It’s telling [AI] what you want it to do, seeing if it did it right, and correcting what it did wrong.”

Some vibe coding tools specifically designed for non-engineers might be risky in the hands of corporate staff like marketers and salespeople, according to some tech execs.

“I’ve got CIOs who are afraid of an app being constructed by an HR person that’s bypassing the security reviews and is going to end up with a privacy issue,” said Tariq Shaukat, chief executive of coding company Sonar.

And while some CIOs are embracing vibe coding, many others are still catching up to the trend. They’re trying to figure out if it’s a new form of “shadow IT,” or a situation where employees bring their own, unsanctioned apps into corporate technology environments.

“A lot of companies I talk to are trying to put in place governance at their audit committee level and things like this, to really make sure it’s governed and controlled,” Shaukat said.

The ever-increasing amount of AI-generated code—and now the arrival of vibe coding—is putting pressure on professional software engineers, tech leaders say, forcing them to think about solving business problems rather than just writing code.

Jude Schramm, CIO of Fifth Third Bank, said the regional bank’s 700 full-time engineers may be entirely vibe coding in a few years’ time.

Schramm said he’s already thinking more about the value of his developers as business problem-solvers rather than as code authors.

“Engineering of the future is going to be much more a world of test, quality assurance and validation,” Schramm said. “That’s what we’re trying to train our folks for.”

Updated: 1-25-2026

Wall Street Has Fallen Out of Love With Software Stocks And In Love With Vibe Coding

Declines show how quickly sentiment can shift as money pours into the next hot investment.

Software companies’ pitch to investors could use an upgrade.

Once a favorite of Wall Street, software stocks have been sliding lately, with investors increasingly concerned about how the sector could be upended by their newest crush: artificial-intelligence companies.

Rocked by the emergence of “vibe coding”—the practice of using AI tools to quickly produce apps and websites—software heavyweights Salesforce, Adobe and ServiceNow all down at least 30% since the start of last year.

An S&P index of small and midsize software stocks is also down more than 20% over that period, with declines accelerating this month after the introduction of Anthropic’s Claude Code, an AI tool that industry insiders have said can dramatically shrink the time it takes to build even complex software.

At a time when many investors are wondering if the AI investment boom itself might be a bubble, software’s slump is a reminder of how quickly fortunes can change on Wall Street.

“The narrative has really shifted,” said Rishi Jaluria, a software analyst at RBC Capital Markets.

Investors, he said, have gone from initially thinking that software companies could benefit from AI to asking, “Is AI just the death of software?”

Investors will gain insight into the overall state of the tech sector in the coming week, with Apple, Meta Platforms and Microsoft set to report earnings.

The Federal Reserve will also meet, though it isn’t expected to change interest rates.

Just a few years ago, it was software, not AI, that was at the center of an investing frenzy.

Over the previous decade, the sector had seemingly made good on Marc Andreessen’s promise that software would come to “eat the world.”

Fueling its rise was the growth of high-speed internet and cloud computing, in which software companies could rent storage space from the likes of Amazon.com rather than own data centers themselves.

Software startups blossomed to serve every niche imaginable, from helping yoga studios handle scheduling and billing issues to securing businesses from cyberattacks.

Once seen as volatile, software came to be viewed by Wall Street as the paragon of stability, because businesses seemed loath to switch products once they became embedded in their workstreams.

Investors also liked the emergence of multiyear subscription contracts that offered steady revenue.

Not only did software stocks soar, the sector became the center of a credit boom, with debt investors eagerly funding a surge of private-equity buyouts.

The pandemic only supercharged the dynamic, thanks to the move to remote work and a drop in interest rates, which made it even easier to borrow money.

But enthusiasm began to wane with the rate increases in 2022 and the shift back to the office.

For debt investors, software companies also began to look less invincible, with competitive pressures and highly leveraged balance sheets driving more businesses into distress.

Before this decade, software defaults were practically unheard of, partly because lending to software companies was so new.

But in the past two years, 13 software companies have defaulted on loans that were broadly syndicated to investors, including both traditional defaults, such as bankruptcies, and out-of-court restructurings known as liability management exercises, according to PitchBook LCD.

One of those companies was Quest, the maker of OneLogin software that authenticates employees logging into their work platforms.

Bought by Clearlake Capital in early 2022 with the help of $3.6 billion in loans from investors, the company benefited from the shift to remote work but was pressured by the weight of its debt and competition from the likes of Okta, a larger, publicly traded company.

It struck a debt-restructuring deal with lenders last June.

Default rates for software loans are still lower than they are for buyout loans generally, and investors have hardly abandoned the sector.

Still, the extra yield, or spread, that investors demand to hold software loans over a benchmark short-term interest rate has ticked higher over the past 15 months, even as overall loan spreads edged lower, according to PitchBook LCD.

“The investor base is definitely scrutinizing these software names much more closely,” said Vince Flanagan, a portfolio manager and senior leveraged-finance research analyst at Seix Investment Advisors.

The emerging threat from AI has only added to that caution. The big risks are that existing software companies could face increased competition from new entrants and that businesses could develop more software themselves rather than paying specialists for it.

In reality, few investors and analysts think that software companies will become obsolete in the foreseeable future.

The more pressing risk is that it could become more difficult to increase revenue, as customers experiment with other options rather than paying more for the usual updates and add-ons, RBC’s Jaluria said.

Ultimately, Jaluria said, AI could damage “fat, lazy incumbents” while helping others that are innovative and can use AI to improve their products.

Uncertainty about the future of software is compounded by the fact that investors lack answers to basic questions about what the emerging AI industry will look like.

Excitement over AI has played a major role in propelling stocks to new records in recent years, with investors especially eager to buy the stocks of so-called AI hyperscalers like Alphabet and Microsoft.

Investors, though, have grown more discerning in recent months in choosing what stocks to buy even among those companies.

As companies start borrowing heavily to fund the build-out of AI infrastructure, debt investors have also taken a more cautious approach than they once did with software companies—forcing big-spending companies like Meta and Oracle to pay high interest rates on new bonds relative to their credit ratings.

“Investor sensitivity is high,” said Rich Gross, a senior fixed-income analyst at Columbia Threadneedle Investments.

Investors, he added, are asking “are these investments sustainable? Are they going to be profitable? Are there going to be cash flows, or will there not be?”

Updated: 2-20-2025

AI Buzzwords You Need To Know Now

To keep up with fast-moving artificial intelligence, learn the terms insiders are throwing around.

As artificial-intelligence research rapidly goes from the lab to the broader world through tools like chatbots, the jargon used by researchers is going along with it.

To understand how AI works in 2025, you’ll need to know what developers mean by “reasoning” and a few other terms that have recently taken off. Here’s our guide to the latest AI buzzwords.

‘Reasoning’

Imagine your boss asks you to arrange an all-hands meeting for the entire department. You check everyone’s calendar to find times when the most people are free.

You check when a large-enough meeting room is available. You match them up and decide when and where the meeting should be.

Congratulations. You’ve done reasoning. AI companies are all racing to get their systems to do the same thing.

Reasoning is a catchall term for the mental work required to solve a multistep problem. Human beings often zip through the steps and call it common sense.

Other problems require more sustained focus over time and might even require backtracking and starting over. Think about planning a multicountry trip. Or developing new theories of quantum physics.

“Reasoning is something we do all the time and sometimes it’s pretty trivial and sometimes it’s pretty complicated,” says Melanie Mitchell, a professor at the Santa Fe Institute who has spent decades studying AI, cognitive science and complex systems.

Such skills have previously eluded AI models. But now that’s changing. OpenAI has introduced two new reasoning models since last fall called o1 and o3 that are now available in ChatGPT (though you have to pay for full access).

DeepSeek’s model that shook up Silicon Valley, called R1, is also a reasoning model.

They can take a minute or longer to answer a question, rather than just a few seconds. But they can solve more complex problems and show you the steps they took along the way.

Reasoning models cost much more than previous AI models to operate because of the amount of computing power it takes to answer each question.

“Whether it’s going to result in human-level intelligence that is going to be economically valuable, I think, is still an open question,” Mitchell says.

—Deepa Seetharaman

‘Evals’

Artificial-intelligence developers are racing to develop tools that surpass human abilities and can work autonomously on behalf of users. But how can we be sure these latest models won’t do harm?

The answer, say AI developers and the governments attempting to oversee them, starts with “evals”—the industry shorthand for evaluations of the kind of AI technology that underlies apps such as the AI chatbot ChatGPT.

Evals are a battery of automated tests that aim to figure out just how capable the latest large AI models are at dangerous tasks such as hacking into a computer network or providing instructions to weaponize anthrax.

The logic is that if the models fail the tests, then they’re safe to deploy. If they pass, more needs to be done to make the models safe.

Most big AI companies conduct evals and promise to limit dangerous capabilities they discover before releasing their models.

Third-party labs and governments are increasingly doing evals, too. The tests often entail a battery of questions about, say, biology-lab techniques, or putting an AI in a virtual environment and asking it to hack a computer or website.

“We’re trying to put in place early tripwires ahead of more advanced capabilities,” says Jade Leung, chief technology officer at the U.K.’s AI Safety Institute, a government body that has run evals on the latest models from companies like OpenAI and Anthropic.

For now, evals are entirely voluntary and reliant on companies wanting to show their tools are safe.

That worries some people who think developers have no incentive to give themselves a failing grade when warranted.

—Sam Schechner

‘Synthetic Data’

During the development phase known as “training,” AI models analyze massive amounts of data to learn everything they know.

AI models typically ingest data created by humans, such as webpages and academic papers. Most advanced models have already consumed every bit of freely available data on the internet, plus whatever proprietary data their developers have access to.

To keep getting smarter, they need more.

Many researchers are now turning to so-called synthetic data—that is, data created by one AI model for another.

AI researchers create synthetic data by instructing a “teacher” large language model to put out data on a topic that a “student” model can use. Those outputs can be a more efficient way to get training data than having humans feed it large papers.

Also researchers can use teachers to teach the student new concepts by feeding pairs of documents and asking it to explain connections between them.

However, the propensity for large language models to make mistakes or “hallucinate” means the outputs have to be carefully monitored by humans. Inaccurate information from a teacher would mean the student is getting dumber, not smarter.

DeepSeek, which upended the AI industry with its latest AI model in January, used synthetic data from one of its models to train another and used open-source technology from Meta Platforms, according to the Chinese company.

OpenAI said it is reviewing indications that DeepSeek trained its new chatbot by repeatedly querying the U.S. company’s AI models.

OpenAI’s terms of service forbid customers from using the outputs of its AI models to help develop their own competitive ones.

—Tom Dotan

‘Vibes’

Tech industry insiders are increasingly relying on intuition, rather than hard data, to judge which AI chatbots are best.

Are a chatbot’s responses concise or overwrought? Pandering or confrontational? Researchers call this a “vibe check,” and to many, it’s just as important as academic benchmarks that measure the ability of AI to answer multiple-choice questions or perform complex tasks.

“Vibe checks are great evals,” OpenAI President Greg Brockman recently wrote on X, using the industry shorthand for evaluations.

A University of California, Berkeley, project named Chatbot Arena has made it simple to perform vibe checks. It gives users answers from two anonymous AI models randomly selected from hundreds submitted by developers.

The question can be anything you would ask a typical large language model chatbot: The user inputs a prompt and gets responses from two models side-by-side.

The user then votes on which answer it likes better. The site, which has received more than two million votes since its 2023 launch, uses those votes to tabulate its rankings, which are closely watched in Silicon Valley.

Researchers are also trying to quantify these vibes. One computer program, dubbed VibeCheck, ranks chatbots along different so-called vibe continuums, such as serious vs. humorous and formal vs. informal.

“It may seem unconventional to focus on vibes instead of concrete metrics of correctness, but these qualitative properties have a measurable impact on how people judge models,” the VibeCheck researchers wrote in a paper.

—Miles Kruppa

Updated: 5-22-2025

Your Next Favorite App? The One You Make Yourself

‘Vibe coding’ with new AI tools makes it possible for just about anyone to create personalized apps—no programming required.

Priya Rose and her friend Irena Wang built the phone app “BabyTime” to allow their young kids to make calls by tapping photos of family members’ faces.

The simple interface is by design, Wang said: Unlike FaceTime, you can’t accidentally call the wrong person. “We don’t want apps full of ads and extraneous stuff.”

While BabyTime is a clever idea, you won’t find it on the App Store. Rose and Wang, who live in New York City, built it in a single afternoon while Rose’s nine-month-old stayed with a nanny.

They copy-pasted code generated by the free version of Claude, Anthropic’s AI chatbot.

Because BabyTime doesn’t need to work for any other families, they “hard-coded” the contact list, baking the photos and phone numbers directly into the app.

And now Wang’s three-year-old is a weekly active user (Rose’s infant is still too young, though she’ll try again soon).

For most of us, software has always been something we consume, not something we create. In the last two years, however, AI code generation products have begun to change that.

General-purpose chatbots like Claude and ChatGPT instantly translate plain English into usable code, though you’ll need to compile and deploy it yourself.

There are AI-powered code editors like Cursor—the fastest-growing software-as-a-service company on record—which is like an autocomplete on steroids for software engineers.

Finally, the most accessible option is “no-code” products like Replit and Lovable, which live in the web browser and are optimized for non-engineers.

You can brainstorm a quick prompt like, “Make me a website to display the new albums I listen to each month,” and within minutes, get a functioning website and URL.

This approach has been dubbed “vibe coding” by Andrej Karpathy, an AI researcher who was on the founding team of OpenAI and led AI at Tesla. AI coding programs “are getting too good. I barely even touch the keyboard,” he wrote on X.

There has been a lot of speculation about how AI coding will help companies to create software in a leaner, faster way—and how it could threaten jobs for junior engineers.

In April, Microsoft CEO Satya Nadella said that up to 30% of code at Microsoft was already written by AI. Meta CEO Mark Zuckerberg expects AI to code at the level of a “midlevel engineer” by the end of the year.

But the most important effect of inexpensive AI coding tools may be to bring software development to the masses, no computer science degree or billion-dollar revenue plan required.

To borrow an analogy from the author and programmer Robin Sloan, AI might enable more people to program like a “home cook”—having fun making apps for their own household and community—instead of relying on fast-food operations that churn out generic apps at industrial scale.

Michael Keating, an e-bike founder in San Francisco, says AI has solved the “iconic nontechnical founder problem”: convincing an engineer to build your crazy idea for you.

He used his $25-a-month Replit subscription to build a game of 20 Questions for his 7-year-old son—“he will play 20 Questions with you until you run out of steam”—and an emoji-typing app he called “Happy Faces” for his 5-year-old daughter.

“I didn’t need to convince anybody else that making a 20 Questions game was worth either my time or theirs,” Keating said. “I just decided it was worth doing, and I went and did it.”

Xavier Maier, an environmental studies student at the University of Waterloo, spends his summers planting thousands of trees for a Canadian reforestation company.

A friend offered to help him make a website to track his planting.

“I didn’t want to make my friend do all this work for free, but he insisted it was super easy,” Maier said.

After “a few hours” of tweaking with AI coding tools, the website is up and running, showing the more than 15,000 trees he’s planted this month.

Hudhayfa Nazoordeen, a friend and former classmate of Maier’s, is a prolific vibe coder: “I probably build an app every day,” he said.

When we jumped on a Zoom, he showed me a series of custom-built apps he made to help study for his exams: a tutorial and calculator for standard actuarial formulas, a “Celestial Mechanics Explorer” to help him understand Lagrange points and black holes through interactive visuals rather than dense texts.

Nazoordeen also does traditional coding—he started building apps from scratch when he was 10—and he says “vibe coding and software engineering feel like two different fields.”

He vibe codes with AI to make “throwaway” apps that he’ll get rid of when they’re no longer useful—for example, once a college class ends.

What current coding agents can’t build on their own is the kind of software that millions of people can use stably and securely.

For commercial software development, expertise still matters. Even to be a vibe coder, it helps to have some familiarity with how software works.

“It’s not as accessible to outsiders as it might seem. Even trying to get a GitHub account was a real pain,” Maier acknowledged. “I have an advantage because my friends are in CS.”

Then there’s the challenge of knowing what to build in the first place. Most of us have designed our lives around using the objects and tools that other people made.

It takes a change of mindset to go from “I wish I had an app for that” to “Why not build it myself?”

Venture capitalist Nikunj Kothari, a 34-year-old based in the Bay Area, is vibe coding up a storm for his own household, creating custom coloring-book sites and meal planning apps.

He muses on app ideas while sitting in traffic. But he’s not sure if vibe coding will achieve mass appeal: “Most people don’t want to maintain software. If they want additional features, they’d rather pay.”

For most purposes, off-the-shelf apps will work just fine. But the last decade has revealed the limitations of Big Tech’s one-size-fits-all model.

Users complain about their favorite websites “enshittifying” as owners jack up prices and fill feeds with ads.

And for groups with fewer numbers or less spending power—for example, visually impaired users, non-English speakers, or yes, babies—it’s harder to find commercial software that fits their needs.

So I can’t help but be excited about the expanded possibilities that vibe coding creates. In my Notes app, I keep an ongoing list of app ideas I want to make with AI.

Some are rote research tasks I’d like to automate away. Others exist somewhere between a creative project and an elaborate joke, like “clickbait headline generator” and “betting markets for my group chats.”

Not a single one is a billion-dollar opportunity. Fortunately for me, I’m the only user-base I need.

Updated: 2-6-2026

Here’s Where AI Is Tearing Through Corporate America

The latest advances have companies—and the stock market—rethinking the software that handles jobs big and small.

In the past few days, employees of Authentic Brands Group have been seeking out the company’s chief digital officer, Adam Kronengold, to ask him a newly urgent question: Should they start using a new AI tool released by Anthropic to review legal documents?

The company, which has about 600 employees and owns brands including Reebok and Champion, uses specialized software for that task now.

But company leaders have encouraged experimentation and embrace of AI in the ranks.

“Everyone feels very empowered to raise their hands and say, ‘Hey, how can we fold this in?’” Kronengold said.

He has told employees he’ll make sure they get access to the new plug-in—but they can keep using the existing software for now.

Envisioning versions of this discussion playing out across corporate America, investors dumped software stocks this week, wreaking widespread losses on companies that sell digital tools and services for human-resources management, productivity, data analysis and other purposes.

Stocks of major software firms such as Salesforce, Intuit and ServiceNow have dropped 10% or more in recent days.

It marked an abrupt acceleration of a trend that’s been playing out for months, with the closely watched IGV software index down around 30% from its late September peak.

The immediate catalyst for this week’s selloff was the release of new capabilities for Anthropic’s Claude Cowork, an AI assistant that lets users assign agents to perform many types of tasks on their computers using only natural-language prompts.

The tools automate workflows and perform tasks across a gamut of job functions with little human input.

The new plug-ins released about a week ago can review legal contracts and perform other industry-specific functions. An update to its model Thursday enhanced capabilities for financial analysis.

The releases appeared to many to confirm a doomsday scenario for the $1.2 trillion software industry, in which companies will be able to “vibe code” their own tools—generating code by plugging natural-language prompts into AI—and cancel their subscriptions.

But the apocalypticism was quickly met with skepticism from software company chiefs, analysts and some executives who oversee technology procurement for their companies.

While few dispute the potential for AI to remake the business-software sector—along with virtually every facet of the economy—most believe the impacts will be more diffuse and delayed than this week’s selloff suggested.

Plenty of incumbents are likely to survive and even thrive as corporate America embraces the wave, they say, particularly those able to adjust their business models to meet the moment.

At a conference in San Francisco Tuesday, Nvidia CEO Jensen Huang waved away the idea that “the software industry is in decline and will be replaced by AI,” calling it, “the most illogical thing in the world.”

“If you were a humanoid robot, would you use a screwdriver or invent a new screwdriver?” he asked.

“Illogical” was the same word JPMorgan analyst Mark Murphy chose to describe, in a note to investors, the “expectation that every company will hereby write and maintain a bespoke product to replace every layer of mission-critical enterprise software they have ever deployed.”

Alphabet CEO Sundar Pichai, too, warned against writing any eulogies: “The companies who are seizing the moment, I think have the same opportunity ahead,” he said on the company’s quarterly earnings call Wednesday.

Among the arguments cited for the continued relevance of software: Vendors have proprietary data and deep industry expertise that’s not easily replicated.

Tech stacks are notoriously hard to disentangle. Many companies view developing their own software as a distraction.

And software purveyors are likely to incorporate every new AI advance into their own products, making them that much more compelling.

Jewelry company Pandora, which sells customizable charm bracelets, uses Claude Code for some of its coding issues, and AI agents handle 60% of customer interactions.

But David Walmsley, Pandora’s chief digital and technology officer, is less willing to rush to experiment when it comes to critical functions such as human resources.

“I’m not going to industrialize my world around a bunch of vibe code,” Walmsley said.

Disaster Potential

Among those cutting some software contracts and turning to AI is the San Diego-based technology company GroWrk.

It has saved roughly $50,000 on an annualized basis by eliminating tools like the project-management platform Asana and others in recent months, and an internal team is now weighing what to build versus buy.

CEO Carlos Escutia said he would hold on to some software essential to the company’s operations, but would increasingly lean on tools like Claude Code to cut out some vendors. “Now you can build these tools internally,” he said. “That’s a good thing.”

At Authentic Brands, an in-house tool powered by models from OpenAI, Anthropic, Google and others allows marketers to generate mock-ups of ads and financial analysts to vet licensing agreements.

It can shave weeks off many projects, said Kronengold, the company’s digital chief.

But in fields such as healthcare, some say, the disaster potential from replacing a tested solution with something hacked together is unacceptably high.

Satheesh Ravala is chief technology officer of Candescent, which makes digital technology used by banks and credit unions.

He has fielded questions from employees about what innovations like Anthropic’s new features mean for the company, and responded by telling them banks rely on the company for software that does exactly what it’s supposed to every time—something AI struggles with.

“If I want to transfer $10,” he said, “it better be $10 not $9.99.”

Anish Acharya, a partner at Andreessen Horowitz involved in enterprise investing, said software costs represent only about 8% to 10% of enterprise spending, a relatively small portion.

Saving a fraction of that with vibe coding isn’t worth the legal and compliance risks for most companies, he said.

One strong sign that vibe-coded software isn’t about to eat incumbents’ lunch wholesale is that OpenAI and Anthropic are themselves customers of numerous large software-as-a-service platforms, points out Aaron Levie, CEO at Box, a cloud-content management firm.

He predicts a flourishing of new custom software that will work with existing platforms.

“If you had a Claude agent go and review a contract, you still need a place to manage the contracts,” Levie said.

A flowering of new options could give software customers leverage in contract renewal negotiations, said Mark Smith, partner at technology research and IT advisory firm ISG.

The added pressure on software vendors could drive a wave of merger and acquisition activity, he said.

The survivors will likely be the ones that pair the advantages of AI technology with their own advantages of scale and expertise.

“Software is not dead,” Ravala said. “The dumb software is dead, or will be at some point.”

Updated: 2-15-2026

Workers Are Afraid AI Will Take Their Jobs. They’re Missing The Bigger Danger

It isn’t whether artificial intelligence is going to replace them. It’s who will control the knowledge that companies capture from their employees.

Walk into any corporate office, and you’ll hear the same anxious conversation: Will AI eliminate white-collar jobs?

The optimists insist that new jobs will emerge to replace the ones we lose—after all, it has happened in previous tech revolutions.

Pragmatists argue the workforce will simply become more productive with artificial intelligence, creating more value with minimal job cuts. And the pessimists fear entry-level knowledge workers will become obsolete altogether.

But this debate misses a crucial dynamic. Right now, workers are potentially training AI how to make them obsolete. And they often don’t realize it.

The kind of AI used by companies, called an enterprise AI system, can capture everything you do at work and use that information to train itself.

These systems can record your interactions within the platform—the prompts you write, the documents you create, the queries you run.

In other words, the company can potentially track—and claim ownership of—every keystroke you make within the system, every idea you document there, every tool you build using that platform.

It can identify what approaches worked best, what email language got responses and how you approached those clients.

And all that knowledge can become part of the company AI, so it may eventually know, down to increasingly fine details, how you do your job.

Then Comes The Dangerous Part For Employees: The AI can pass that information along to anybody else who does your job, or in some cases just do the job itself.

Over time, you could become a lot less valuable to your employer—and a lot more replaceable.

This dynamic may fundamentally change the relationship between employer and employee. The stakes are so high and so urgent that both sides are rushing to position (or protect) themselves.

Executives are rapidly implementing enterprise AI systems, seeking productivity gains and competitive advantage—and they often aren’t disclosing the implications for job security and privacy.

Meanwhile, at least some employees are secretly adopting personal AI tools, sometimes violating corporate policies, so that their employers can’t capture everything they know and do.

Capturing The Essence

To understand what’s coming, you need to understand what enterprise AI systems actually are. These are different from the interfaces you use at home.

Enterprise AI systems are platforms that integrate directly into corporate workflows—think of Microsoft Copilot embedded in Word, Excel and Outlook, or Salesforce’s Einstein AI woven into customer-relationship management.

These systems sit inside the tools where you already work. And they can potentially capture much of your work within the platform, learning from many interactions, and embedding that knowledge into company-owned infrastructure.

What once lived only in employees’ heads, built through years of experience and hard-won expertise, is increasingly being institutionalized in real time.

When you leave, at least some of your knowledge stays behind, embedded in systems that will be used by the AI and by your replacement (if a replacement is needed at all).

Imagine that you’re a senior software engineer debugging a system crash. You run a bunch of tests to figure out the problem, and when you discover the solution isn’t in the documentation, you develop a novel workaround.

You share the solution with the company, obviously, but the expertise and techniques that you brought to the problem were all yours, in a fundamental way. You figured out the workaround because of what you know and how you work.

That is the way things used to be, anyway. When you do your work through enterprise AI, though, the system doesn’t just record your solution.

It Can Capture Your Problem-solving Approach: which questions you asked first, how you refined the search when initial attempts failed, potentially even the logical steps that led you from symptom to cause.

The next time junior engineers face a similar crash, the system may be able to guide them through elements of the methodology you used.

You haven’t lost your expertise. But now the employer also has access to key aspects of that expertise, in a form it controls and can deploy to other employees without you.

It has a partial blueprint for how you think, and some of the knowledge that once made you indispensable is now a reproducible company asset.

Making It Personal

These revolutionary changes seem to put workers in a tight spot. But I believe employees have an alternative—one that isn’t easy, but could help move the power dynamic back in their favor.

Specifically: Employees should consider avoiding their company AI systems when possible and use personal AI tools like ChatGPT, Claude, Gemini, Copilot or dozens of others.

These tools operate on completely different terms than enterprise AI. You access them directly. You own your prompts, your workflows, your customizations.

The knowledge you create stays with you. Most critically, when you walk out the door, your AI-enhanced capabilities walk with you.

Maybe you’re required to use your company’s enterprise AI for client work. But all the strategic thinking you do before engaging with clients? Develop that using personal AI tools.

I spoke with a regional vice president at an energy company who does exactly that: He uses his firm’s enterprise system for required compliance and documentation, but develops new analytical approaches and tests complex decisions in personal AI tools. The novel insights stay his.

What Can Be Done?

Using personal AI tools is just the first step employees should take, however. To really change the power dynamic, they can act on other fronts.

• Negotiate Upfront. When joining a company, people should treat access to AI tools like intellectual-property ownership.

Most employment agreements cover IP created on the job, but employees should dig further into a company’s policies before signing on: What gets captured through enterprise AI?

How long is that data retained? Can you use personal AI tools for skill development? Can you request deletion of your contributions if you leave?

Most companies haven’t thought through these questions yet, which means there is room to establish reasonable boundaries before you’re locked in.

• Support Collective Action. Individual opt-out of AI is often impossible, so unions and professional associations need to pay attention. With collective bargaining, workers could demand transparency about the use of enterprise AI and demand fair compensation for the knowledge it gathers.

Without collective power, individual employees will keep clicking “accept” on agreements that restructure their jobs simply because they have no alternative.

Concerted employee action may start to change the AI calculus. Employers may find that enterprise AI systems do capture knowledge, but at a steep cost: They may drive away the most talented employees, ones who realize they can build more valuable, portable capabilities with personal tools.

AI is breaking the traditional model of employment in real time faster than anyone realizes. The companies and employees who understand these dynamics will position themselves to capture AI’s benefits.

Those who don’t may find themselves on the losing side of the biggest workplace shift in a generation.

Bitcoin Information & Resources (#GotBitcoin)

Ultimate Resource On Baby Formula Shortage

Houston Software Executive Charged With Largest Tax Evasion Case In History Against An Individual

The Best Solo Miners. Feat. The NerdQaxe++

Bitcoin Bulls Take Over The Military

The Three High-Tech Steps That Can Lower Your Energy Bills

Sauerkraut, Kimchi And…Gut Nuts? Fermented Foods Are Having A Moment

America’s Car Repossessions Business Is Booming If You Have What It Takes!!

These AI Power Users Are Impressing Bosses And Leaving Co-Workers In The Dust

11 Most Famous Black Female Artists

Monty: Is Job-Hoarding Officially Over? AI: Yes