Thinking in public.

At Ampliflyer, we're obviously vested in making the future of AI work for businesses. But we're just as vested in understanding the day-to-day business of AI through an informed, strategic, and data-driven lens.

So, when we heard about OpenAI's latest joint venture pitch to the private equity community, we thought it might be helpful, interesting, and admittedly amusing to get the opinion of StreetAnalyst, one of our own AI-native business intelligence applications. We provided the link and simply asked the question, "What does this deal signal about OpenAI?" What follows is directly from our application—research, charts, opinions, snark and all.

Enjoy,
Justin Daab, CEO, Ampliflyer

The 17.5% Solution: OpenAI's High-Stakes Hustle

The Smell of Napalm and Preferred Returns

In the kingdom of venture capital, there is a fine line between a "visionary leap" and a "desperate scramble." When a company is growing revenue at 233% year-over-year, you typically expect them to be dictating terms to investors from a throne made of silicon and hubris. But OpenAI's latest overture to the private equity titans—TPG, Advent International, and Bain Capital—smells less like a victory lap and more like a high-octane refueling mid-flight.

OpenAI is reportedly dangling a 17.5% guaranteed minimum return to lure these PE shops into a joint venture. To put that in perspective, even the most aggressive hedge funds would blush at guaranteeing 17.5%. This isn't a sign of strength. It's a sign that OpenAI's capital appetite has outpaced even the most generous venture capital buffet, and now they need institutional money with deeper pockets and longer time horizons.

The Most Expensive Party in Human History

Let's be blunt: OpenAI is burning cash at a rate that would make a Pentagon procurement officer nervous. The company is projected to spend $20 billion in 2025 alone. Revenue? Around $12.7 billion, meaning the net burn rate is roughly $7.3 billion per year. That is not a "growth phase." That is a "we need to build a Death Star and we need to build it yesterday" phase.

But here's the thing—OpenAI's user base is staggering. 900 million weekly active users. ChatGPT's growth trajectory has defied even the most optimistic projections. And compute costs? They're running at roughly $17 billion annually. So the question is whether the revenue can catch up to the burn.

Scaling at the Speed of Light (and Debt)

A Moat Built of Gigawatts?

Perhaps the most underappreciated argument for OpenAI's dominance lies in its sheer physical footprint. In 2023, OpenAI's compute capacity was a respectable 0.2 gigawatts. By 2025, that number exploded to 1.9 gigawatts. That is a nearly 10-fold increase in two years. To put that in context, 1.9 gigawatts can power nearly 1.5 million homes. OpenAI isn't just a LLM provider; they are becoming one of the largest industrial consumers of energy and compute on the planet.

While OpenAI has the users (62.5% market share), the quality of their "moat" might be more porous than the valuation suggests, especially as Google's Gemini grows at 30% month-over-month compared to ChatGPT's 3%.

The Arithmetic of Artificial Ambition

The IPO Bridge: The 17.5% guarantee makes sense if OpenAI is planning a 2026–2027 IPO. PE firms will want a clear exit, and a public offering is the most liquid path. The guarantee is essentially a "bridge loan" sweetener to buy time until the public markets take over the funding burden.

The Enterprise Yield: OpenAI's enterprise revenue is growing faster than consumer, and with significantly better margins. The PE capital will likely be funneled into sales infrastructure and the Stargate partnership to capture more of the enterprise market.

The Cash Flow Goalpost: OpenAI has publicly stated they expect to be cash-flow positive by 2029. That's four years away—an eternity in tech. The 17.5% is the "tax" OpenAI is willing to pay to survive until 2029.

The Great Re-Rating

The popular view is that OpenAI is the inevitable winner of the AI era. My analysis suggests something more volatile.

The Bold Prediction: OpenAI will successfully raise this capital, and the PE firms will get their 17.5%, but it will come at the cost of OpenAI's soul—and its cap table. By 2027, the "guaranteed return" will be seen as the moment the music slowed down. While OpenAI will remain the dominant consumer brand (the "Google" of AI), the real profit will migrate to the "boring" players who figured out how to make AI profitable at a smaller, more efficient scale.

And one day, Wall Street will ask the uncomfortable question: "Was 1.9 gigawatts of compute ever really a moat, or was it just the world's most expensive electricity bill?" Meanwhile, the PE firms will smile, cash their 17.5%, and move on to the next deal—leaving the public markets to figure out who's holding the bag when the music finally stops. The PE firms will walk away with their 17.5% "guaranteed" pound of flesh, while retail investors will be left holding a bag of "potential" that costs $17 billion a year to maintain.

Justin Daab, CEO, Ampliflyer

This week, Harvard Business Review published an article, "Researchers Asked LLMs for Strategic Advice. They Got 'Trendslop' in Return." in which the authors, Angelo Romasanta, Llewellyn D.W. Thomas and Natalia Levina asked, "...as LLMs are integrated into executive workflows, a critical question emerges: How good is their advice? Is it trustworthy?"

The article summarizes their conclusion:
"...leading LLMs have clear biases when it comes to strategy and consistently recommend strategies that align with modern managerial buzzwords and trends rather than context-specific strategic logic. This propensity for AI to opt for buzzy ideas over reasoned solutions is called 'trendslop,' and leaders should beware of it warping their strategic planning. When using AI in strategic planning, leaders should: use it to expand options, not make choices; counteract known and potential biases; remain alert to changing biases; watch out for the hybrid trap; and not rely on context alone."

It's not a surprising result. We've all seen it. You ask ChatGPT for strategic advice and get back something that sounds like it was cribbed from a TED Talk — all "differentiation," and "long-term vision," and "collaborative ecosystems." It sounds smart. It sounds polished. And it's the same advice it gives everyone else, regardless of whether you're running a construction company in Toledo or a fintech startup in Berlin. That's "trendslop." And the researchers behind this HBR piece ran thousands of simulations to prove it's not a fluke, and that it's baked into the models.

I would, however, propose another question the authors may have also wanted to ask: is there a smarter way to access and harness the knowledge in these models that might deliver unique insight and sound, tailored advice? Spoiler alert: Ampliflyer started addressing this problem almost two years ago and our technology was designed from the outset to do just that.

They asked the LLM to choose. Never let the LLM choose.

The researchers gave five leading LLMs a set of MBA-style scenarios and asked them to recommend one of Porter's generic strategies. The result was a massive overweighting of "differentiation" across every model. But here's the thing: that outcome was baked in the moment they asked the question that way. LLMs are statistical pattern machines. When you ask a generalist model "what strategy should this company pursue?" you are asking it to sample from a distribution that is overwhelmingly shaped by internet discourse—LinkedIn posts, Medium articles, HBR itself—where "differentiation" and "innovation" vastly outnumber discussions of cost leadership or focus.

Defined personas, roles, procedures, and motivations aren't cosmetic. They're structural constraints.

The researchers prompted LLMs as generic "strategic advisors." That's the configuration where trendslop is worst—the model reaches into the broadest, buzzword-heaviest slice of its training data. Every agent in every application Ampliflyer builds has not only roles and actions they are designed for, but we also add extensive mindset instruction for each task. A forensic audit agent, for instance, is told it's a skeptical and meticulous forensic accountant where precision, accuracy, and intellectual honesty are non-negotiable.

The authors walked right into the totally avoidable "hybrid trap".

When the researchers saw a blend of differentiation and cost leadership in a response, they coded it as a "hybrid strategy." But a model combining elements of both isn't evidence of strategic thinking. It's evidence of averaging. That's an important distinction. Without structural enforcement—like requiring explicit trade-off justification or scoring each strategy on independent axes—LLMs will naturally blend outputs because blending is, statistically, the safest response. In our applications, we build structured analytical flows where each strategy dimension is evaluated independently, with its own evidence requirements and explicit trade-off surfaces.

Want a smarter answer? Create structural disagreement.

One of the more elegant findings in the paper is that multi-model ensembles slightly reduced bias. That's a start—but it's the weakest form of a much more powerful principle: productive friction. At Ampliflyer, our applications don't just query multiple models; they instantiate structurally opposed analytical perspectives within the same workflow.

You want Porter back into the room? You have to invite him.

The researchers noted that LLMs failed to consistently apply the strategic frameworks. In our suite of applications, analytical frameworks aren't suggestions—they are procedural scaffolding. When an Ampliflyer application agent runs, it doesn't ask "what strategy should we pursue?" It walks through the framework's own analytical structure, step by step. This means that in the analysis cost leadership always gets equal structural weight to differentiation—not because the LLM was convinced it's a good idea, but because Porter's work is encoded as a first-class framework in the system.

Humans should stay in the loop. By design, not by accident.

The researchers found that better prompting barely moved the needle—roughly a 2% shift on the strongest biases. Adding rich organizational context only shifted things by about 11%. In many of our applications, Ampliflyer implements what we call checkpoint-based execution. After the structuring phase generates its hypotheses and selected agents, the engine stops and waits for the human to approve, modify, or redirect. The LLM proposes; the human decides whether it's the right scaffolding. The LLM executes and provides the raw analysis and a structured synthesis of what the results were along with an auditable trail of why it arrived at that result.

The problem was never the model.

The technology worked as designed. The problem was the approach to using the technology and the expectation we've all been sold, that the genie in that bottle is smart, reasoning, and an omniscient expert in everything, out of the box. The problem was the absence of structure. It was asking a concept calculator to perform a judgment that requires wisdom, experience, and genuine understanding of trade-offs—and then being disappointed when it returned the most popular answer instead.

Garbage in, garbage out; precision in, precision out. The researchers proved the garbage-in half of that equation. Ampliflyer demonstrates what's possible when you take the precision-in half seriously. And the difference, both in the quality and the usefulness of what comes back, isn't incremental. It's transformational.

—Justin Daab, CEO, Ampliflyer

Generative AI is a powerful tool for efficiency and innovation, but ...

The recent, high-profile security incident of the reported breach involving McKinsey's Lilli platform has exposed a stark reality: functionality without real trust is an existential risk.

For CIOs, CISOs, and legal teams, this incident was a potent validation of their greatest fears, underscoring that the "move fast and break things" mentality is fundamentally incompatible with enterprise AI that processes sensitive data.

Why Ampliflyer started with the core concept "Privacy by Design".

The market is now demanding 'zero-knowledge' AI architectures where data privacy is an intrinsic architectural principle, not an afterthought. That's why Ampliflyer explicitly designed our platform around implementing principles that directly mitigate these risks:

• Client-Side Processing: All sensitive user inputs, attachments, and data prep happen locally in the user's browser, eliminating the transmission of raw data to our servers.
• Absolute Non-Visibility: Ampliflyer is structurally prevented from seeing customer data. We don't see your IP, nothing goes into a database, and there is nothing to breach because we literally do not have any customer data to steal.
• Enterprise-Grade Guarantees: Leveraging enterprise APIs with stringent contracts that prohibit data use for model training, adding another layer of verifiable trust.

Think of trust as currency; local processing and non-visibility are about building the foundational trust needed for broader AI adoption in highly sensitive sectors like finance and healthcare.

The future and ultimate value of enterprise AI will be defined by its trustworthiness.

Many companies are struggling with getting AI tools past compliance; and, with many of the available tools, those compliance concerns are justified. However, to accelerate the process of finding a compliance-friendly set of tools for your business, focus your thinking around these core three lenses:

• Redefine AI Procurement: Demand verifiable Privacy-by-Design platforms and scrutinize data handling contracts. Prioritize audited security architectures and certifications.
• Invest in AI Security Literacy: Implement mandatory training across your organization on the unique risks of GenAI (like prompt injection) and foster a "Secure by Design" culture.
• Establish Robust Governance: Map your AI data flows, implement strict access controls, and form cross-functional AI Governance Boards to monitor evolving regulations.

The era of trust-infused AI has arrived. This is the moment to transform the challenge of AI security into a distinct strategic advantage, fostering an environment where strategy, intelligence and growth thrive on a bedrock of unwavering trust.

– Justin Daab is the Cofounder and CEO of Ampliflyer.ai, an AI-Native Platform for Strategy and Intelligence used by Next-Generation Leaders.

Unlocking breakthrough potential by leveraging purpose-built AI applications, without breaking the bank.

Executive Summary

The global economy is undergoing a profound transformation driven by Artificial Intelligence. While large enterprises with extensive capital and R&D budgets have historically spearheaded AI adoption, a unique and potent opportunity is now emerging for the middle market. Often defined as companies with annual revenues between $50 million and $1 billion, these firms are the economic backbone of many nations but frequently face significant capital constraints that limit their access to cutting-edge technology.

This thought leadership piece argues that the rise of "AI-native" applications—technologies fundamentally designed around AI rather than having it as an add-on feature—offers a transformative pathway for these capital-constrained middle-market companies to achieve not just incremental, but exponential growth. By leveraging purpose-built AI solutions, often available as purpose-built SaaS applications, middle-market firms can unlock capabilities previously reserved for Fortune 500 companies: from hyper-personalized customer engagement and predictive supply chain optimization to accelerated product development and sophisticated data-driven decision-making.

This piece explores how these converging forces create compelling synergies and unprecedented opportunities, while also candidly addressing the challenges and potential disruptions that middle-market leaders must navigate. It provides a clear, actionable Blueprint for Action encompassing strategic vision, data governance, phased adoption, innovative talent models, and proactive risk management, empowering business leaders to harness AI-native technology and propel their organizations into a new era of competitive advantage and sustainable growth.

The Blueprint for Action

1. Develop a Clear Strategic Vision and Foster Leadership Alignment — Educate leadership, align AI goals with core business objectives, and secure buy-in from the top down.
2. Prioritize Data Foundation and Governance — Assess data readiness, invest in data quality and integration, and establish clear data governance policies.
3. Adopt Incrementally with a Focus on High-ROI Use Cases — Start with pilot projects, scale what works, and measure everything rigorously.
4. Develop a Smart Talent and Organizational Strategy — Invest in upskilling your current workforce, outsource and partner strategically, and build an "AI Translator" role.
5. Proactive Risk Management and Ethical Governance — Establish AI governance frameworks, prioritize cybersecurity, and plan for workforce transition.

Read the full report on LinkedIn →

Kelley Buchanan joins Ampliflyer's Executive Advisory Board—from early user to power user. Kelley's customer testimonial says it all:

"I was initially skeptical, having seen countless 'hype pitches' for revolutionary technologies fail to deliver. But Ampliflyer's suite is NOT hype—the promise of 'months to minutes' is real. These powerful, AI-native solutions, including DesignThinker, DisruptionCompass, and StrataGem, are true game-changers. They put the scarcest resource—time—back into the hands of executives, delivering quality, executable insights in minutes, not months and millions spent on standard consulting engagements. Now we have the latitude to explore every angle, deeply and creatively before committing to a single direction. I recommend Ampliflyer to serious executives who are looking to accelerate outcomes and generate the smartest solutions."

—Kelley Buchanan, Executive Advisor, Board Member, and Former Strategy Partner, Fortune 100 Senior Strategy Officer, Senior Financial Officer and General Manager

Ampliflyer is transforming the work of strategy, product, innovation, and corporate development leadership. Are you leaning into your future?

Justin Daab, CEO, Ampliflyer

Generally, we are biologically wired to avoid failure. Failures can come at great cost. But for innovators and entrepreneurs, failure is an investment. If you walk into any Silicon Valley coffee shop you could toss a Ferrari key fob in any direction and probably hit a freshly-minted billionaire bragging about how many times they crashed and burned before they bought their private island. Don't misunderstand here, "Fail Fast" isn't some masochistic initiation rite for tech founders. It's a proven economic strategy designed to maximize learning while minimizing the exposure in cost and time. It's about knowing what's working, what needs fixing and, most importantly, what might never work.

That being said, even among the most devoted members of the cult of Fail Fast, "fast" was always relative. It at least meant weeks, sometimes months, of grinding out business plans, refining pitch decks, and honing market analyses before you even had a snowball's chance of getting meaningful feedback—let alone funding.

From Napkin Sketch to Pitch Deck, Right on Your iPhone.

VentureVision is an AI-native business plan and pitch deck generator, built from the ground up to run on iOS. It's not a chatbot bolted onto a document template. It's a purpose-built engine that takes a raw business idea and generates a comprehensive, investor-ready business plan and pitch deck—complete with market sizing, competitive analysis, financial projections, and strategic positioning—in the time it takes to finish a coffee. It is, by design, the antithesis of the months-long consulting engagement, providing a quicker path to finding product-market fit or, indeed, failing spectacularly and moving on.

Your Ideas and IP Stay Yours. Always.

Crucially, in an era where data privacy is no longer a luxury but a fundamental expectation, Ampliflyer emphasizes data privacy. VentureVision is the perfect example. You can see it right in the App Store listing "This company collects no user data." We never see your idea. Your concept inputs are not used to train any LLM model. We believe this isn't just good practice; it's a critical competitive differentiator for a tool that asks users to entrust it with nascent, often confidential, business concepts.

Success, or Sunk Costs, Anyone Can Afford.

VentureVision consistently delivers high-quality, nuanced reports that genuinely accelerate the learning process (not just the document generation) for less than the price of the coffee you'll finish while it runs. It democratizes access to high-quality strategic planning, leveling the playing field for entrepreneurs who lack the resources for traditional consulting. It's like having a team of analysts in your pocket, available whenever inspiration strikes.

Put a Unicorn in your Pocket.

VentureVision is available right now on the App Store for iPhone and iPad.

We released our first commercial product!

Meet VentureVision. Available for iOS and live on the App Store.

Move from that business idea you've been dreaming about to your first plan and pitch deck in minutes not months.