If you only have six months of experience in an industry, you usually aren’t equipped to poach your employer鈥檚 clients and start a competing business. Of course, you may believe otherwise if you plan to use your employer鈥檚 proprietary information to train an AI agent to do the heavy lifting for you.

That is exactly what the defendant attempted in I litigated on behalf of a client in the U.S. District Court for the Northern District of California. In that case, the defendant, a former senior employee at the marketing agency, didn鈥檛 just copy client lists to solicit accounts. Instead, he , attempting a shortcut to create a competing business without spending years developing the skills and expertise typically needed to provide such services.

AI is creating new ways to misappropriate trade secrets, well beyond the classic theft story of an employee downloading files onto a thumb drive. This case highlights a new reality for founders, exposing unprecedented risks related to AI and trade secrets, employee use of personal devices, and the heightened need for companies to protect themselves in the age of agentic AI.

Key lessons for founders

AI contamination cannot be undone

The defense in this case argued that because their client placed the stolen documents in a “quarantine folder” and no clients had been poached yet, no harm had been done.

The court appropriately recognized that the 鈥渜uarantine folder鈥� was grossly insufficient, granting our temporary restraining order and preliminary injunction while recognizing the competing AI agent as an “existential threat.”

An agreement not to touch just the stolen documents themselves is essentially irrelevant once AI gets involved: If a trade secret is used to develop a machine learning model or used to train an AI agent with recursive capabilities, it contaminates the model and the impact can be difficult to undo. Even if raw source files are completely scrubbed of any direct evidence of stolen materials, defendants will struggle to prove to courts that the AI isn’t operating a certain way precisely because it once had access to that proprietary information.

Personal devices can pose big risks听

While trade secret misappropriation has typically been malicious, AI now also creates significant risks of inadvertent violations. Startups frequently scale with employees working on personal devices under standard Confidential Information and Invention Assignment Agreements. Running desktop AI agents on these devices can pose a serious risk.

Unlike standard web interfaces, a desktop AI agent can scan files, create folders and ingest directory paths across a local hard drive, often without requiring special permission or alerting the user.

Right now, thousands of employees may be inadvertently violating confidentiality agreements by running consumer AI agents on personal laptops that house their employers’ IP. If settings allow model training from this data, courts may eventually find these trade secrets to have lost legal protection due to dissemination outside the company.

Standard IT audits won’t catch AI data theft

This recent case serves as a warning for how attorneys and founders must approach trade secret investigations. Traditional digital discovery is no longer sufficient because AI tools make it effortless to disperse, alter and recreate data across nonobvious folder paths.

Actions that previously seemed like overkill 鈥� such as full forensic imaging of a personal device 鈥� are now the bare minimum. Furthermore, AI dramatically reduces the friction of reverse-engineering software and tech products, potentially removing telltale signs that code or data was copied.

Proving theft in AI contexts requires intensive, highly technical expert analysis and carries far more uncertainty than traditional IP litigation.

3 practical steps companies should take

Founders should take three legal and operational steps to proactively protect their business:

1. Implement strict agentic AI policies: Mandate tight controls and settings that isolate outside AI from company documents. Web-based tools like Claude with strict data privacy policies are vastly safer than open source or desktop apps that crawl local drives. Some AI agents should not be permitted to run at all within an environment containing trade secrets.
2. Eliminate remote access from unmanaged devices: Shift to a closed cloud environment where data can be viewed and manipulated on a secure server, but files cannot be downloaded locally to a personal desktop. Removing local file downloads eliminates the risk of desktop AI agents indexing proprietary data.
3. Rewrite your Confidential Information and Invention Assignment Agreements for the AI era: Standard confidentiality agreements need explicit clauses addressing AI use, model training inputs, and a precise definition of data misappropriation through machine learning.

If you suspect an employee has walked away with data or misused it in an AI tool, your first step should be immediate: cut off all system access and retain counsel. Don鈥檛 wait until you start losing clients or experiencing financial losses. Feeding trade secrets into AI presents an existential threat from day one.

AI gives individuals the unprecedented ability to compete in spaces where they lack experience, making proprietary data more vulnerable than ever. Modernizing security, policy and legal strategies to address today鈥檚 AI threats will help startups and established companies avoid catastrophic exposure.

is a partner at , where he focuses on complex commercial litigation, including disputes related to trade secret and IP misappropriation, corporate governance and breach of fiduciary duties, breach of contract, and other complex business disputes. He specializes in preventing and resolving all manner of disputes among and between corporate founders, directors, officers and shareholders, LLC members, and other current and former business partners.

Illustration:

When a company delays its IPO to show investors better numbers, that’s a strategy. When the NYC pension fund publicly asks you to delay it over AI safety concerns, your chief rival’s CEO calls his own IPO “ill-advised” and postpones it until at least next year, tech stocks are sliding on open-weight model fears, and your own prospectus warns of “existential risks to humanity” from the product you’re selling, that’s not a strategy.

That’s a market sending you a message.

‘s numbers are extraordinary. Its third-quarter 2026 revenue was $16.8 billion 鈥� up 45% quarter on quarter. That means the company鈥檚 annualized revenue run rate is roughly $67 billion.

What the S-1 actually says

Here are the numbers from Anthropic’s :

• $4.6 billion: Anthropic’s full-year 2025 revenue.
• $7.33 billion: What Anthropic spent on compute and infrastructure in 2025 alone 鈥� more than its total revenue.
• $42 billion: Anthropic鈥檚 net loss in 2025. The roadshow will immediately explain that $34 billion of it was a noncash accounting charge 鈥� an adjustment to the estimated value of financing instruments that could eventually convert into shares.
• $518 billion: How much Anthropic expects to spend, at least, over the course of a decade on AI infrastructure with six partners, including , and , according to the confidential IPO prospectus by . About 80% of the total is noncancelable or payable regardless of usage.
• $2 trillion: That鈥檚 the IPO valuation Anthropic is still targeting for November, a figure representing 435x its 2025 revenue.

One more number from the S-1 that the roadshow will bury: A quarter of Anthropic’s 2025 revenue came from two unnamed customers.

Three signals the market is sending

The IPO delay is the headline. It isn’t the only signal worth reading.

Signal one: NYC Comptroller Mark Levine. Last week, the supervisor of New York City’s pension funds 鈥� which hold roughly $7 million in Anthropic exposure today, but could hold dramatically more if Anthropic enters the Russell 3000 post-IPO 鈥� sent to Anthropic CEO urging the company to pause the IPO entirely.

The risks he cited are the same ones Anthropic’s own prospectus flags: increasingly autonomous AI systems that could behave in unexpected ways, create security problems, and in the document’s own language, pose “existential risks to humanity.”

Signal two: delays 鈥檚 IPO. OpenAI has said it will not go public in 2026. CEO Altman called an IPO at this moment given AI safety concerns, and while that framing is partly strategic 鈥� positioning OpenAI as the more responsible actor 鈥� it鈥檚 also a real signal when the CEO of the second-most valuable AI company in the world looks at current conditions and decides to wait.

Signal three: The chip selloff. Last week鈥檚 was driven partly by concern over AI-buildout costs. 鈥檚 helps explain the concern: Through 2030, it sees at most $3.6 trillion in revenue attributable to AI data centers, versus $8.5 trillion needed to cover planned investment, financing costs and a 12% return.

The open-weight problem

Here is the structural argument that no amount of Q3 revenue growth resolves: Anthropic’s $2 trillion valuation requires closed-model AI to maintain a durable pricing premium over open-weight alternatives. That premium is being systematically dismantled.

Open-weight AI is challenging the economics behind closed-model valuations. Enterprise adoption rose from 23% to 67% in 2026, with companies reporting 70%鈥�90% cost savings. R1 offers frontier-level performance on most enterprise benchmarks at 7 cents per million input tokens; independent analysis puts Anthropic鈥檚 comparable input and output pricing at roughly 18x and 23x higher, respectively.

DeepSeek鈥檚 weights are free to download, allowing enterprises to self-host without a model-provider API contract or per-token billing, while the capability gap with closed models reportedly closes within six months of a frontier release.

Against that backdrop, 鈥� the research firm that called Anthropic鈥檚 proposed offering 鈥渁n unprecedented test of investor gullibility鈥� 鈥� that justifying a $2 trillion valuation would require an immediate 10% after-tax operating margin sustained for a decade, extraordinary revenue growth, and $358 billion in annual after-tax operating profit by 2035.

Why this is worse than 1999

The dot-com bubble had a genuine excuse. Nobody knew what the internet would become. The AI IPO wave of 2026 does not have that excuse.

Worldwide spending on AI is , up 47% year over year, according to Gartner. Yet a July 2025 estimated that 95% of the organizations it examined were getting no return from generative AI. Meanwhile, said in its earlier this year that just 7% of surveyed leaders say they鈥檝e established ROI from AI.

This is the context in which Anthropic is asking investors to pay 435x 2025 revenue for a company with $518 billion in locked infrastructure commitments, 25% revenue concentration in two unnamed clients, and a pricing moat that open-weight models are closing in six-month cycles.

The delay from October to November doesn’t change any of those numbers. It adds one quarter of revenue data.

What actually survives this correction

Here is what no AI IPO roadshow will say out loud: The biggest enterprise AI winners will not necessarily own the most expensive model. They will own the architecture that turns any model into governed, auditable, measurable outcomes 鈥� through proprietary enterprise data, domain-specific fine-tuning, multiagent orchestration and controls that make autonomous execution trustworthy.

That value does not depend on keeping model capabilities scarce or securing a $2 trillion valuation. It compounds in production, where customers understand that the model is not the product. The deployment is.

The dot-com crash did not happen because the internet was bad. It happened because the price was wrong. AI investors face the same distinction: Real technology does not automatically justify its valuation. A strong quarter and a one-month delay do not close that gap. November simply gives Anthropic more time to explain the math.

, the co-founder and CEO of , is an entrepreneur and product leader with more than 20 years of experience across cybersecurity and AI. He has built three generative AI companies, including one that reached unicorn status. Gurtu is known for turning advanced AI into scalable, real-world platforms, and previously contributed to successful exits, including acquisitions by , and , and a public company, . He holds a master鈥檚 degree from the and CISSP certification.

Related 小蓝视频色情网页版 queries:

Illustration:

You鈥檝e just moved into a new house. It鈥檚 unfurnished; the walls are bare. Ten years ago, you might have loaded up and started searching for furniture that you like, consulting interior design magazines and looking for offers on retailers’ websites. In 2026, however, there鈥檚 an entirely different option at your fingertips. You could take a photo of the empty room, upload the image to an AI agent, share your budget and style, and get a list of furniture that suits your preferences. You could even have the agent purchase on your behalf.

Instead of just selling to one customer base 鈥� humans 鈥� brands are now selling to two.

This second audience of AI agents makes decisions on different criteria than humans. Often, in fact, AI agents will make purchasing decisions before a human ever sees the product page. There鈥檚 a real chance that all the emotive copy marketers have so carefully crafted will get broken down and reconstituted by a machine. Just relying on the human touch isn鈥檛 enough anymore.

Leave humans in the equation

A word of caution, though: Marketers shouldn鈥檛 forget the human entirely in the rush to prepare for AI agents. It鈥檚 easy to get carried away with visions of an entirely agentic future. But at the end of the day, each agent is still catering to a human with their own preferences, and personalization will only get more granular over time. The trick is to optimize for both audiences simultaneously, not just one.

Newer startup founders have the advantage here. Established brands have to retrofit decades of copy, guidelines and approval processes to make themselves agent-ready. New brands, on the other hand, can build in copy and a brand voice that鈥檚 legible to humans and agents from the very start.

But what does this look like in practice?

To start, brands should work out a distinct voice 鈥� and write it down. In most early-stage companies, the brand voice solely exists in the founder鈥檚 head. This might work if the copy鈥檚 written by a handful of people who work with the founder, but it doesn鈥檛 work well when the writer 鈥� or reader 鈥� is a machine.

AI tools follow explicit rules and drift without them. I’ve spent years at my startup, , working out how to get AI agents to write better content, but it’s only now that we’re seeing agents on the other side of the process too 鈥� and that makes the rules matter even more.

What a brand sounds like matters, but what it doesn鈥檛 sound like matters just as much. Brands should note words never to use and the claims never to make. Often these instincts don鈥檛 get written down; they feel obvious to the team who already know them. But if you take the time to spell them out before you have a team 鈥� or a machine 鈥� producing your copy, you鈥檒l save yourself from running into issues further down the line.

Verify your brand

Marketers also need to realize that product descriptions aren’t just administrative: they鈥檙e now core brand assets. In an agent-mediated world, they may be the most-read brand copy you own, since agents read them when looking for a product to recommend or purchase. If two personalized AI agents look for furniture for two different people, they may surface the same chair for both, but with the description tweaked for each user. The source description needs to carry the brand intact.

Where possible, brands need to test channels for both readers. Email inboxes are the clearest current example of the 鈥渟econd reader.鈥� Both Gmail and now summarize messages before the recipient opens them and offer summaries above the body of the email 鈥� which may be read before the email itself.

So before sending, check what the AI summary says your email is about. Does the offer survive? Does it maintain what makes you unique? It鈥檚 not time-intensive, and it鈥檚 a small habit any marketing team can adopt today with no tooling.

Brands also need to ensure they鈥檙e measuring the outcome of their marketing 鈥� not just the volume. Marketers who measure the success of their materials only by the amount produced will find their data meaningless as AI makes content production easier and easier. Focus instead on business outcomes, conversions and revenue. These are the metrics that will show if your marketing team is really doing their job.

And finally, don鈥檛 lose sight of the human. No one knows how fast AI purchasing will scale. Humans are still predominantly the ones reading copy and making purchasing decisions. Even AI agents, the second audience, cater to their users鈥� preferences.

New founders and brands are best positioned to take advantage of the transition to a dual-audience economy. Though it鈥檚 hard to say exactly how, the retail landscape will look very different in five or 10 years. Building these decisions into your brand now is easy.

Retrofitting them later is expensive.

is chief product and growth officer at, an AI platform that helps brands generate and optimize bespoke messaging at scale.

Illustration:

Acquisitions are reshaping the technology industry. Venture-backed AI companies acquired 195 AI startups this year through late September, already exceeding the total for all of 2025, 小蓝视频色情网页版 data shows.

For investors and founders, such acquisitions can represent major financial milestones. For employees holding equity compensation, a transaction can also change personal financial plans they had expected to follow for years.

A client whose employer was acquired in an all-cash transaction described to me the atmosphere inside the company during the weeks that followed: Employees repeatedly discussed the deal in office corridors, trying to understand what it meant for their equity awards, their jobs and their future. She compared the confusion to the atmosphere surrounding a major reorganization or a round of layoffs.

The transaction offered the prospect of substantial financial gains. Yet employees were struggling to establish what those gains would mean for them individually.

One acquisition, different outcomes

Employees at the same company can emerge from an acquisition with remarkably different financial outcomes.

Some own vested shares. Others hold restricted stock units, stock options or performance-based awards. Their rights depend on grant agreements, vesting conditions and transaction terms. Even two employees holding RSUs may face different payment schedules or employment requirements.

Depending on the agreements, outstanding awards may be settled in cash, accelerated, converted into replacement awards, or remain subject to future conditions. One employee might receive a substantial payment when the transaction closes. Another may have much of the expected value tied to continued employment.

The announced acquisition price provides a reference point. It rarely tells an individual employee everything they need to know about the money they will receive.

This helps explain the conversations my client described. Colleagues may compare the value of their equity without realizing that their contractual rights differ.

Wealth and employment uncertainty

Technology employees with substantial equity compensation often depend on their employer in several ways.

Their salaries, bonuses, career opportunities and investment portfolios may all be linked to the same company.

An all-cash acquisition can reduce their exposure to its share price once vested shares are converted into cash. Other forms of dependence may continue.

The acquiring company may reorganize teams, change responsibilities or introduce retention arrangements. Employees whose unvested awards remain conditional on employment may face difficult choices about staying.

Someone expecting a large payment can therefore feel financially stronger while becoming less certain about their career.

Before making plans

A major acquisition announcement can change personal ambitions long before the proceeds arrive.

Employees may begin considering early retirement, a new business, a career break or a major family purchase.

Those possibilities are worth exploring. But an announced acquisition may still require regulatory approvals and other closing conditions. The headline value of an equity award may differ substantially from the cash ultimately available to its holder.

Taxes, exercise costs where applicable, outstanding liabilities and payment schedules can all affect the outcome. So can the employee’s financial position outside the company.

A substantial payment might make a career change affordable without providing enough capital to fund permanent retirement. Existing pension savings, household spending and future earning capacity remain important.

In my work with technology executives, I have seen expectations about new wealth reshape personal plans before the underlying numbers have been established.

Employees facing a cash acquisition can begin with three questions: What rights do my specific awards provide? When will the proceeds actually become available? And what choices will the resulting after-tax capital support?

The answers may differ considerably among colleagues.

An exit without an exit decision

Founders who sell their companies often participate directly in negotiating the transaction. Employees holding equity compensation generally have much less influence over its timing.

They may have intended to retain their shares for years, sell gradually or remain employed until additional awards vested. A corporate acquisition can replace those plans with a timetable established by the transaction.

For some employees, the result will be substantial financial flexibility. For others, a significant part of their expected compensation may remain connected to future employment.

The period between announcement and closing can be especially difficult because the financial opportunity is visible while important details remain unresolved.

My client’s description of the office corridors captured that experience. The company had made a decision about its future. Its employees were left to work out what that decision meant for theirs.

The acquisition determines the terms of a corporate transaction. Each employee still has to establish how much financial freedom it creates.

is co-CEO and managing partner of , which works with technology executives, founders and families on integrated wealth planning, equity compensation and major liquidity events. His background includes more than two decades in the technology industry.

Disclosure: The views expressed are the author’s own. This article is for general informational purposes and does not constitute financial, investment, legal or tax advice. Details of the client experience have been presented without identifying information.

Related 小蓝视频色情网页版 queries:

Related reading:

• 小蓝视频色情网页版 Data Shows AI鈥檚 Most Active Startups Are Becoming Serial Acquirers
• AI Is Creating Wealth Faster Than Financial Lives Can Adapt

Illustration:

Today鈥檚 venture capital ecosystem looks markedly different from even just five years ago.

Most significantly, a small number of AI companies are now attracting the vast majority of available venture capital. In Q1 2026, just four companies collectively raised $188 billion, which alone accounted for 65% of total global venture investment, per 小蓝视频色情网页版 data.

Meanwhile, in that same quarter, just 20% of global VC funding went to non-AI firms. meaning that checks are being made out to fewer companies all around.

So what happens to startups when the VC route is all but closed?

I built a company into a multibillion-dollar business without any venture funding. Here鈥檚 a few lessons I picked up along the way:

Prioritize customers and revenue, not fundraising

Many founders unknowingly treat fundraising as a measure of success. But while it鈥檚 useful, funding doesn鈥檛 result in success on its own.

Achieving strong levels of customer trust is also helpful when it comes to inspiring investor confidence. High retention rates, repeat purchases and sustained engagement drive the听 financial fundamentals on which strong valuations are built. In B2B SaaS, for instance, companies with net revenue retention above 110% . Additionally, customers provide a feedback loop that helps create the discipline every company needs, and each user is a vote of confidence that compounds over time.

Going to market: Take time on pricing strategy

Too many early-stage companies set prices too low because they fear that higher prices will repel customers. In my experience, however, underpricing more often signals uncertainty rather than good value. Customers want quality products and services, and they expect that the quality will be reflected in the price.

Incredibly, , and firms more commonly regret underpricing than overpricing. So at the outset, it’s important for startups to inform themselves what price their product or service can actually sustain, and then raise quality if it鈥檚 too low.

Equally, in services markets where prices are more established around a norm, superior quality can be an important differentiator. At the same time, however, companies must continually resist the temptation to chase customers with ever-cheaper prices, which can start a race to the bottom.

Hire the best talent

Too many founders hire for speed rather than quality. Take the time needed to raise the bar with talent. Hire the right people, and they鈥檒l help maximize every last drop of value.

Many years ago, I decided to focus on just a small fraction of our applicants and . When we found the best developers, it meant we could compete on delivery quality rather than just price. Keeping costs low is important, but skimping on talent is not a winning strategy.

Use process automation efficiencies to drive growth

One of our best decisions was to at an early stage. Not only do these systems screen developer applications, but they can also build, match and deploy top-quality teams for each client project.

This advance allowed us to cut costs associated with evaluating applications, accelerate lead times, and improve the quality of our teamwork. In parallel, so that growth wouldn’t require proportional increases in overhead.

Whether it鈥檚 in recruitment, sales or operations, process automation can create some of the biggest drivers of growth and help bring about conditions needed for significant expansion to occur.

Operate as if cash is running out tomorrow

Companies thrive in challenging funding environments by operating with discipline. That means obsessing over unit economics, prioritizing activities that directly drive revenue, questioning every cost, and eliminating anything that isn’t pulling its weight.

Not only does this aid growth, but it also makes a business more resilient in hard times. Financial constraints drive leaders to create efficiencies by automating workflows, streamlining processes and inventing innovative work models.

Years before COVID-19 forced others to follow our lead, our company streamlined its operations and boosted scalability by developing a remote-working model. When the pandemic made remote work necessary, the fact that we already had the systems and workflows in place gave us a competitive advantage over other firms, and this in turn led to significant growth.

Long term: Stay married to the problem, not the solution

The past decade convinced many founders that venture capital was the default path to building a meaningful company. But while funding cycles rise and fall, the fundamental principles of good business don’t. Infusions of capital can accelerate growth, but they cannot create a great company from scratch.

On the longest timescale, the most successful founders are the ones committed to solving problems rather than just selling a product and chasing funding. In a world short on funding, growth simply can鈥檛 happen without it.

is the founding GP of , and chairman of , one of the largest bootstrapped nearshoring companies worldwide. Driven by his entrepreneurial spirit, he launched BDev Ventures in 2021 to help B2B companies accelerate their growth through a proprietary lead-generation platform. He also invests in and serves on the boards of biotechnology and e-learning companies. In recognition of his career path, he won distinctions such as an EY Entrepreneur Of The Year 2024 Bay Area Award.

Related reading:

• Q1 2026 Shatters Venture Funding Records As AI Boom Pushes Startup Investment To $300B听

Illustration:

Congratulations to the enterprise AI company that just raised another nine-figure round to solve a problem software is supposed to solve by itself: getting humans to use the product.

You know the one. Won鈥檛 name it. Don鈥檛 have to. You鈥檝e read the press release 鈥� “autonomous,” “agentic,” “category-defining” in paragraph one, and then, a few paragraphs down, where the money is actually going: hiring an army of engineers to sit inside customer offices and manually make the “autonomous” thing behave.

Nothing says 鈥渨e鈥檝e built the future of work鈥� quite like raising half a billion dollars to hire more people to do the work.

The business model, decoded

Here鈥檚 how it works, in case the deck was too artfully vague:

1. Sell the customer an “AI platform.”
2. Discover the platform doesn鈥檛 actually do the thing on its own.
3. Send in the forward deployed engineers: sharp young people with CS degrees who now spend six months hand-building the workflow the website implied already existed.
4. Bill the customer for “software.”
5. Book it as ARR.
6. Raise a bigger round because 鈥淎RR go up.鈥�
7. Repeat, forever, with an increasingly large and increasingly expensive engineering org standing between the product and the promise it made.

This is not a flywheel. This is a payroll with a pitch deck.

And the industry has stopped even pretending otherwise 鈥� it鈥檚 fundraising on it, loudly.

pledged valued at $10 billion, with another $4 billion from ,, , and . The vehicle, DeployCo, has been with more than $4 billion in investment.

Meanwhile, job postings for “forward deployed engineer” are up somewhere between 800% and 1,000% year over year. Lightcast data reported by from January to August 2026. data carried by put listings up more than 800% from January to September 2025, and data from 643 in April 2025 to 5,330 in April 2026. Total comp is pushing past $300,000 in the U.S. list at $280,000 to $320,000 base pay and OpenAI at $185,000 to $300,000 base, plus equity. Fortune reported some Anthropic roles reaching $400,000 in total compensation.

Somewhere, a VC is calling this 鈥渙perating leverage鈥� with a completely straight face.

It is not operating leverage. It鈥檚 a staffing agency that learned to say 鈥渁gentic.鈥�

The tell

Ask the one question the pitch deck conveniently never answers: 鈥淚f you pulled the embedded engineer out tomorrow, does the customer still get what they paid for?鈥�

If the honest answer is 鈥渨ell, no, but the ARR is real鈥� 鈥� that鈥檚 not product-market fit. That鈥檚 people-market fit, and people-market fit has a brutally unglamorous ceiling: You can only hire humans so fast, and you can never hire them cheap enough to earn a software multiple. Every new logo doesn鈥檛 get more profitable to serve. It gets another headcount request.

Two companies, same $20 million ARR. One needs 10 engineers to keep the lights on. The other needs 100. Both slides say, 鈥淎I software company.鈥� The difference shows up in gross margin, cost to onboard the next customer, and whether that cost falls as the company scales.

The bill always comes due

Here鈥檚 the thing about people-market fit: it feels great right up until someone asks the unfashionable questions: What does it actually cost, fully loaded, to onboard the next customer? Does that number go down as the company scales, or does it just grow the org chart? What happens to gross margin if the first year growth slows and the headcount can鈥檛 hide behind hypergrowth anymore?

Companies funding their FDE armies with fresh nine-figure rounds are betting nobody asks these questions until after the round closes and the confetti is swept up. That鈥檚 a fine bet in a hype cycle. It鈥檚 a terrible one in a market that鈥檚 about to start reading the fine print.

Because eventually, someone always reads the fine print. A CFO builds a cost-per-deployment model. A CIO benchmarks time-to-value against three vendors instead of trusting one demo. An investor finally does the unglamorous work of separating software ARR from human-powered ARR on the same balance sheet.

The distinction buyers and investors can test is on the numbers. Software ARR should get cheaper to deliver as it scales. Human-powered ARR does not. OpenAI guaranteed the private-equity backers of its deployment vehicle an annual 17.5% return over five years, which is a different structure from a growth thesis.鹿

So no, your AI isn鈥檛 autonomous. But your business model might be 鈥� autonomously transferring the cost of a services company onto a software valuation, for as long as the music plays.

Enjoy the round. Hope the org chart was worth it.

, the co-founder and CEO of , is an entrepreneur and product leader with more than 20 years of experience across cybersecurity and AI. He has built three generative AI companies, including one that reached unicorn status. Gurtu is known for turning advanced AI into scalable, real-world platforms, and previously contributed to successful exits, including acquisitions by , and , and a public company, . He holds a master’s degree from the and CISSP certification.

Illustration:

How would you handle a junior employee questioning the way you鈥檝e always done something? Might you roll your eyes and dismiss their comments as naivety? Or would you indulge their curiosity?

As AI has made knowledge accessible on a scale never seen before, companies are becoming leaner and operating faster. What once took a large team months to do can now be done by a handful of people in a few days.

It is undeniable that autonomous systems are changing the way we work. In the small and dynamic teams of the future, choosing the right humans to direct your AI systems will become make or break for a company.

As a six-time founder, I鈥檝e learned the best workplaces are the ones where questioning things is the norm. Here鈥檚 how to build a team that questions orders as much as it obeys them.

AI by design

鈥淎I by design鈥� doesn鈥檛 only mean building the right autonomous systems. It鈥檚 also about creating autonomous humans. You need to trust the human in the loop as much as you do the technology. The team around you ultimately shapes the company in ways you can鈥檛 control. So surrounding yourself with people you trust to make decisions when you鈥檙e not in the room is critical to success.

The nature of job roles is changing. A job description can be totally different from the one advertised by the time you鈥檝e found a candidate, and a six-month roadmap is outdated by the time you鈥檝e finished drawing it up.

The has stated that as AI reshapes the workplace. Hiring for a fixed set of skills won鈥檛 move the needle. The people worth investing in are those who are able to learn what comes next.

Hire for curiosity

Curiosity is the best trait you can hire for. When I look at a CV, it鈥檚 not just experience or academic credentials that draw me in. I look for what people have built on the side. What have they made on their own initiative?

Hire for curiosity, and you鈥檒l create an environment where people are willing to throw new ideas around and challenge outdated ways of doing things. If your team is worried about saying the wrong thing, they won鈥檛 end up saying the right thing either.

In a world where technology is turning everything on its head, the companies that will thrive are the ones with curious people who are able to adapt to change quickly.

Hand over real responsibility

Employees with autonomy operate faster. They make better decisions because company vision is not abstract to people who are given real responsibility to drive it. Being involved in strategic direction stitches an employee into the fabric of the company and keeps them engaged.

Trust your employees to know what isn鈥檛 working. Allow people to pursue new ideas and reinvent things. Your employees are likely to spot potential problems before you do when they鈥檙e closer to the product. Make sure your team has the autonomy to try to fix them before they take hold.

Boredom is a big business risk

As companies become smaller, employee retention is more important than ever. Losing your best people is costlier when you have fewer of them to begin with.

The effects of a bored employee surface quickly in a small team. Boredom makes talented people a little bit slower, and a lot less likely to try the difficult thing.

The challenge for modern leaders is keeping your team engaged and interested. A bored employee is more likely to be swayed by a flattering recruiter.

Make sure to invest in development, keep your team learning, and always encourage curiosity. Give people autonomy and responsibility from the beginning, and boredom has nowhere to take hold. The mentality of risk and the drive to challenge something outdated is what you need to succeed in the changing work environment.

Dare to be a beginner, and teach your team to do the same. That junior employee questioning how you’ve always done things could be the most valuable person in the room.

is co-founder and CEO of Stockholm-based , an agent-driven, game-based learning platform. He has experience building, exiting and scaling digital companies across 70 countries. Uniplay is his sixth company to date.

Illustration:

A first venture usually follows a well-worn playbook: a gap in the market, obsessive refinement of concept and pitch, eventual conviction that the solution is as bulletproof as it is revolutionary.

Then, the dreaded phase of product-market fit. Instead of building the future of an entire industry, you find yourself forced to focus on a narrow use case and spend the first invaluable years proving that your idea can make money and scale. The grand vision gets put on pause.

For most founders, this journey is the only one available. After all, investors are betting on someone who has never built a company before, so a small check and a narrow product limit what everyone stands to lose. But some ambitions need total commitment from the start, and having built a company before helps when taking them on.

Lessons from the front line

After selling my first venture, , to for 鈧�1.8 billion (about $1.9 billion USD) in 2022, I found myself in a new position 鈥� and it wasn’t horizontal on a sun lounger. My co-founder and I had a real war chest of our own, and a brand-new entrepreneurial bug. For the first time in our lives, we had the experience and the resources to undertake something truly world-changing.

Compliance is a gauntlet that any founder will be painfully familiar with. Scaling Tink across Europe, to 700 people operating under a host of regulators, my team and I got used to tearing our hair out over small wording tweaks, reviews that dragged on for months, and perpetual uncertainty that stopped us from iterating fast and boldly. Products were often delayed by months, a slip that meant the world in an industry moving as fast as ours.

A problem with too many heads

Compliance was a problem that could never be solved piecemeal. Every company’s obligations depend on a million different factors, each of which sits in a different part of the business that rarely talks to one another.

Legal tracks regulations; another department oversees how they are implemented. Product teams make their own calls on features, and yet another team opens a case to check whether it was compliant. Policies, standards and regulations belong to other departments entirely, and those are usually the part of a silo a compliance startup builds for 鈥� a clear buyer and a budget.

“Solving compliance” was a mission akin to fighting the Hydra: cut off one head and two more grow back. There is no minimum viable product here. Fixing one head changes nothing. What it takes is a maximum viable product from day one.

Why the second time is different

In recent years, we’ve seen a series of inspiring examples of successful founders coming back to take on new problems. Stockholm knows this well: alumni of and have gone on to found and back a long list of new companies, using the credibility from their first successes to try something harder.

The second time around, capital and access both come easier. A successful first exit opens doors to talent, partners and investors that we simply didn’t have with Tink. It is a fantastic opportunity, and one I’m grateful for. You get to bring every lesson from what went well and what went badly, and you start with a real tailwind when people want to work with you and back you.

Money also means you can fund the full build yourself, so no investor asks you to shrink the idea before you’ve tested it.

Maximum viable product

Experience and capital used to get you only so far. Ten years ago, building something this broad would still have needed hundreds of people of various backgrounds, and many years. Language models have changed that. Our team today is 22 strong and growing, and ships at a pace I could only have dreamed of when we first built Tink.

We also feel we’ve grown thicker skin from the bruises of our first 15 years as entrepreneurs.

You know the setbacks are coming, and you know that a venture is never a one-man sprint but a team marathon. None of that guarantees success. But it’s a good place to start if you want to stand your ground and build your biggest vision yet.

is the CEO and co-founder of , the Stockholm-based AI-native compliance company that came out of stealth in September 2026. He previously co-founded the open banking platform in 2012 and led it as CEO through its growth to a multimarket operation with around 700 employees before its 鈧�1.8 billion acquisition by in 2022.

Illustration:

You have likely never heard of . You should have.

Over the past decade, Tosin and his team built into the operating system for more than 6 million Nigerian businesses, , or .

As a Nigerian American, I鈥檝e witnessed my father make trips to Western Union to send money back home for decades, and I experienced Nigeria鈥檚 fee-laden payment system when visiting the country. Moniepoint is transforming that system.

Moniepoint is not alone. , , , , 听1 鈥� these are companies most Americans have never heard of, despite them being worth several tens or hundreds of billions, following the trail blazed by , and .

These are super apps that bundle banking, commerce, ride-hailing and bill pay into one trusted application. The idea never took hold in the U.S. for one reason: trust is cheap here. With , a specialist app inherits the credibility of the financial system. Where trust is cheap, there is no need for a super app.

Build trust first

In emerging markets, trust is the whole problem. Four conditions tend to precede a super app: High mobile penetration, which makes distribution cheap; weak incumbent competition; poor existing financial services; and the condition most investors screen past, a financial crisis severe enough to strip incumbents of legitimacy. These populations are not merely underbanked. They were harmed, and the institutions that harmed them still hold the market.

Brazil froze roughly 80% of bank deposits and financial investments for 18 months in 1990 in an effort to break hyperinflation. The banking consolidation that followed left five banks and . That was the void Nubank was born into.

Kazakhstan went through it twice: savings erased by inflation in the thousands of percent when the ruble zone collapsed in 1993, then a 2007鈥�09 banking crisis in which the country’s largest bank, BTA, holding roughly a quarter of the system’s assets, Kaspi grew into that void.

Southeast Asia鈥檚 1997 crisis was harsher still. Indonesia closed 16 banks in a single announcement, the %. Two decades later which enabled Sea to build and into that gap.

Kaspi also answers the objection I hear most from American investors, which is that these markets are too small to produce large exits. Kazakhstan has only 20 million people, and according to company , around 14 million 鈥� nearly 70% of its population 鈥� use Kaspi. The . It earns . Depth of spend, not population, sets the ceiling.

Compare Nubank to . Nubank has 14x the customers, to , about 7x the revenue and a market capitalization more than 6x larger. Nubank . A technology applied in an overlooked market with a large unaddressed pain point can build a bigger and more defensible company than the same technology applied in a rich one.

Screen for the four conditions and a few markets stand out as places where the next super app could emerge:

• Bangladesh: , with the government now promising to pay 7.5 million savers whose money is gone.
• Egypt: Repeated devaluations of the pound between 2016 and 2024, shrinking savings held at banks the state itself owns.
• Vietnam: after its owner was arrested in a $24 billion fraud.
• Pakistan: , then government-built rails and a national ID that makes onboarding nearly costless.

One boundary condition: The crisis must be severe enough to delegitimize incumbents but contained enough that the currency and payment rails still work.

The first generation of super apps created enormous value for the investors who backed them early. In 2001, the South African media company paid roughly $34 million for a 46.5% stake in Tencent, and . 鈥檚 investment in Nubank鈥檚 $2 million seed round, combined with its lead Series A investment, was worth $7 billion at the company鈥檚 IPO.

The second generation is scaling now. The third will be built on stablecoins and AI technologies, in Abidjan, Buenos Aires, Karachi and Lagos, by founders like Tosin Eniolorunda who will play an increasingly important role in the growth of emerging market economies.

听 is co-founder and managing partner of , where he oversees the firm鈥檚 investing strategy, governance and compliance, with a focus on fintech and enterprise technology. Since co-founding 645 Ventures in 2014, he has invested in and partnered with companies including , , , , and . Previously, he was an investor at , where he invested in 19 companies and had nine exits representing more than $9 billion in exit value.

Illustration:

In 2021, putting AI on a slide could get you funded. Small and mid-sized businesses were eager to be early adopters, and they bought fast. I watched companies secure customers, raise money, then struggle to keep anyone using their products. Retention fell off a cliff.

Many of those products were generic: They did what traditional software already did, with AI fairy dust sprinkled on top. Others saved a bit of time on something nobody’s job depended on. A 20% improvement on a noncritical task makes for a nice demo and a dead renewal.

There were never easy AI wins, even if there was easy money going around. The market is now correcting for that. Companies delivering meaningful value are growing faster than anything I have seen in enterprise software. For investors, the question is which of these companies can keep their customers 鈥� and whether the coming IPOs will expose the difference.

What makes an AI business defensible today

Before my co-founder and I wrote a single line of code, we interviewed more than 900 mechanical engineers, from juniors to VPs. We asked where their hours went and which of those hours they would pay to get back. Those 900 conversations shaped how I think about building an AI business.

First: the improvement has to justify the investment. Saving 10% on an occasional task will rarely change how a company operates. Reducing a business-critical process from weeks to minutes, or saving serious money, will.

Second: domain expertise. Frontier models are becoming infrastructure. Access to a good LLM is available to almost anyone with an API. The commercial opportunity lies in understanding an industry well enough to solve problems that generic tools cannot.

Third: proprietary context. In terms of engineering examples, every company that builds physical products has accumulated decades of engineering knowledge. Much of it remains buried in old drawings or locked away in the heads of experienced employees. That knowledge is absent from a foundation model’s training data. An AI product that cannot access it will struggle to become essential for the business. The advantage grows as the product becomes embedded in the customer鈥檚 workflows and replacing it becomes costly.

What this means for funding and IPOs

The upcoming AI IPOs will put these business models under much greater scrutiny. Public investors will examine net revenue retention and gross margins, assessing whether rapid growth is able to translate into a sustainable business. They will also want to know whether each deployment becomes cheaper to support as the company grows. The standards they establish will influence what private investors expect from earlier-stage AI companies.

I can already see the change in my own fundraising conversations. A few years ago, $1 million in ARR was a milestone. In the AI era, a new product can reach $1 million in ARR in about a year, and shut down a year later. Now ARR alone no longer tells investors enough.

What investors keep asking about is expansion. They want to know whether customers are increasing their spending after the initial deployment. Expansion is the closest thing we have to proof that a product has changed how an organization works: customers have experienced its value and committed more of their own budget to it.

Over the next 12 to 24 months, I expect capital to keep flowing to companies that combine deep industry expertise with significant improvements in business-critical work and access to customers’ proprietary knowledge. They may grow through hands-on deployment rather than viral self-service, but they can expand inside accounts that trust them.

Companies whose products are thin layers over someone else’s model, with impressive customer lists and retention charts nobody wants to show, will find their next funding round much harder than the last one. I would rather grow fast and steady for five years than spectacularly for five quarters.

is the founder and CEO of , the first AI for mechanical engineering 鈥� a large mechanical model for physical product design. He conducted AI and mechanical engineering research at as a Fulbright postdoctoral fellow and became the youngest Ph.D. graduate in the history of the Farid has built a community of more than 60,000 engineers and supports underserved youth through his nonprofit initiative.

Related 小蓝视频色情网页版 query:

Illustration: