Copyright forbes
A Fortune 500 CEO pulled me aside after a board presentation last month. "We keep talking about AI strategy," he said, "but I'm starting to think we're solving the wrong problem. When customers stop coming to our website and just ask ChatGPT what they need, do we even have a business anymore?" He's asking the right question at exactly the wrong time. Because the platform inversion he's worried about? It already happened. Applications no longer contain AI. AI now contains applications. That architectural shift fundamentally changes who controls customer relationships, how businesses compete, and where value gets captured. And if you're still debating whether to "adopt AI," you've already lost the game. The Diagram That Explains Everything Last week, Sam Altman unveiled OpenAI’s corporate restructuring from nonprofit to public benefit corporation (PBC). But the real revelation wasn’t the legal paperwork - it was the architecture diagram showing how OpenAI sees every business fitting into their future. At the foundation: chips, data centers, energy. The compute power - the physical infrastructure. Next layer: the frontier AI models they build. The intelligence layer (that eventually becomes Artificial General Intelligence (AGI) and then Artificial Super Intelligence (ASI). Above that: the OpenAI Account. A single identity controlling access to everything. And then there are applications and devices. These will all use the subsequent layers (that can be described as their “platform”). Some of these will be by OpenAI, such as ChatGPT, Sora, their new browser Atlas, and eventually their own proprietary hardware devices. Inevitably, their offering will all be powered by the OpenAI “stack” and the company will vertically integrate software, hardware, and services. Think Apple's ecosystem, but with your customers' conversations, preferences, and purchasing history as the core asset. MORE FOR YOU And then there’s the last box labeled "Third-party applications." That's you. That's every business. You're categorized as an add-on to their system. OpenAI shared their expansive vision of the future as a platform where people and companies build on their AI cloud infrastructure. Notice what's missing? Your direct relationship with customers. In this architecture, businesses don’t own customer interactions - they rent algorithmic consideration. When someone asks ChatGPT for a restaurant recommendation, software tool, or consultant, OpenAI's system decides which options to surface. You're no longer competing for customer attention. You're competing for an algorithm's permission to exist. You're not a partner in this system. You're inventory. The Transfer Of Power Hiding In Plain Sight Applications no longer contain AI. AI now contains applications. it’s a profound statement and explains the entire power shift. And if you don’t grasp what it means, you’ll miss how radically the business landscape is changing. For 30 years, you built the container. You created the website, designed the app, opened the store. Customers came to your space, on your terms. If you wanted AI capabilities, you added them to your platform - a chatbot on your site, recommendations in your app, automation in your workflow. You controlled the environment. You owned the customer relationship. That just inverted. Now customers live inside AI platforms. And your business gets called into their conversations - if the algorithm decides you're relevant. It seems like every week there are new bricks being laid down as the new AI foundation is being built right before our eyes. Every partnership announcement, every technical breakthrough, every infrastructure commitment makes the inversion more permanent. In October, OpenAI announced that apps like Booking.com, Spotify, and Zillow can now operate directly inside ChatGPT conversations. Users never leave the platform. Over the past two months, OpenAI has announced commerce partnerships with Shopify (connecting their entire merchant network), Walmart (enabling direct purchasing through chat), and PayPal (integrating millions of merchants so customers can discover, compare, and purchase without ever touching your website). Sam Altman shares his vision for apps inside ChatGPT at OpenAI's DevDay 2025. This isn't about APIs and SDKs. It's about a transfer of power. What happens when customers don't visit websites, call salespeople, or walk into locations? What happens when they simply describe what they want to AI, and it handles everything else? Your business just moved inside someone else’s platform, whether you agreed to it or not. Unlimited Memory Changes Everything The last constraint that might have prevented total platform dominance just disappeared: memory limitations. For decades, computing systems forgot - every session started fresh, creating opportunities for businesses to build relationships because platforms couldn't remember what you learned face-to-face. That advantage just evaporated. As former Google CEO Eric Schmidt observed, we're accelerating toward "infinite context windows" that can "remember everything," and we're already most of the way there: OpenAI's GPT-5 handles 400,000 tokens (roughly 800 pages), Anthropic's Claude reaches 1 million tokens in beta, Google's Gemini operates at 1 million today with 2 million planned, and xAI's Grok runs at 2 million tokens now. So what happens when your application is in the AI platform? Examples of how apps like Booking.com, Expedia, Spotify & Zillow show up in ChatGPT. When customers ask AI platforms for recommendations in your category, the platform knows their complete commercial history - every product researched, every complaint filed, every preference expressed, every purchase made - while you know what they bought from you last quarter. AI platforms will know their decision-making patterns, budget constraints, quality preferences, and unspoken needs across their entire life, and uses that complete profile to recommend whatever option best serves the customer - including your competitors when they're the better fit. This creates a compounding moat: more usage generates more data, better data produces better recommendations, better recommendations drive more reliance, and more reliance reduces the need for direct business relationships. Even if you’re "integrated" into these platforms, you’re competing for recommendations based on customer insights you can’t see, against selection criteria you don’t control, using algorithms that optimize for platform objectives, not yours. The platform doesn’t just sit between you and your customer - it becomes the customer’s memory, advisor, and procurement agent all in one, and when customers trust AI platforms with their complete commercial history, the switching costs trap you, not them, because the customer's entire decision-making context lives in the platform's memory, not in your CRM. Your relationship with customers becomes AI-mediated. The platform owns the interaction. You execute the transaction - if the algorithm selects you. The $125 Billion Dependency Trap OpenAI projects reaching $125 billion in revenue by 2029, according to internal forecasts shared with investors. That’s approaching Nvidia’s current scale. But OpenAI isn't selling chips or ads. They're selling access to customers. The company expects $29 billion from AI agents alone by 2029 - autonomous systems that research options, compare providers, and complete transactions. These agents will operate at premium pricing: $2,000-$20,000 monthly for enterprise services. Here’s the math that should terrify you: Despite 800 million weekly ChatGPT users, OpenAI won't reach positive cash flow until 2029. It projects burning roughly $8 billion annually. Those economics require alternative monetization: transaction fees, data licensing, enterprise deployments, and perhaps taking a percentage of commerce happening within the platform. When AI platforms decide they need 15% of transactions flowing through their interface, what's your negotiating position? You have none. When billions of transactions flow through a handful of platforms, you accept the terms or opt out of where customers make decisions. What Dies When AI Owns Your Customer Discovery disappears. You can't optimize for search rankings or buy ads when the selection happens algorithmically inside an AI conversation you're not part of. Google's Gemini processes up to 2 million tokens (roughly 4,000 pages). It can read your entire web presence and decide you're irrelevant to the customer's query. You'll never know why. Differentiation compresses. When AI presents three recommendations based on price, availability, and generic ratings, your unique value proposition becomes bullet points in an algorithm-generated list. A boutique hotel owner told me: "We spent years curating experiences. Now ChatGPT describes us in the same format as Holiday Inn Express. How do we compete on bullet points?" You don't. That's the problem. Pricing power evaporates. When AI presents options with prices displayed prominently, competition defaults to price. Service quality, brand reputation, and customer relationships matter less than being the cheapest option the algorithm recommends. This leads to another derivative effect that will prove existential: When AI platforms control customer relationships and mediate all transactions, they accumulate data about every business in every category. They see which providers get selected, which get rejected, what factors drive decisions, where pricing breaks, and how customers really behave. That’s not just data. That's market-making power. The $1.4 Trillion Moat You Can't Cross In the internet era, business sought to build their own websites and apps, but it didn’t make any economic sense to build out your own hosting infrastructure once cloud computing platforms like Amazon’s AWS, Microsoft’s Azure and Google Cloud took off. Now we’re seeing something similar happen in the AI era, as AI infrastructure requirements create permanent barriers. The combination of cost-prohibitive compute power combine with your applications and services now being offered within the AI platforms is going to be a shock to every business. OpenAI committed approximately $1.4 trillion to data center capacity, targeting 30 gigawatts with plans to add 1 gigawatt weekly at $20 billion per gigawatt. Google's Vertex AI usage grew 20x year-over-year. Anthropic's Claude maintains focus on complex tasks for 30+ hours. Only a handful of tech companies can afford to compete as frontier AI providers: OpenAI (Microsoft-backed), Google, Anthropic (Amazon-backed) and Elon Musk’s xAI. Sure, others are vying to compete such as Meta, Salesforce, Apple, etc. However, “AI is a sport of kings,” as D.A. Davidson analyst Gil Luria stated. Only a few companies will control the interface layer for all commerce, services, and information. Your business exists inside their platforms - on their terms. Everyone will become dependent on their infrastructure. This isn’t a temporary competitive advantage. It’s a restructuring of economic power at the infrastructure level. When a few platforms control how billions of people access products and services, that’s not innovation. That’s market consolidation disguised as technological progress. The Fork That Offers No Good Path Every business faces the same decision tree: Path One: Integration. Get listed in ChatGPT, Gemini, Claude or Grok. Optimize for AI recommendation algorithms. Accept that customer relationships flow through intermediaries who control access and take a percentage. This offers survival. You’re part of conversations where decisions happen. But you’re also dependent. When platform economics shift - and they will - you have little leverage. Path Two: Resistance. Maintain direct customer relationships. Build brand value that transcends AI recommendations. Invest in differentiation algorithms can't capture. This preserves independence and control over pricing, positioning, and customer experience. But you also watch billions of customer conversations happen in spaces you're excluded from. Neither path is sustainable. Integration makes you dependent on platforms optimizing for their outcomes, not yours. Resistance excludes you from where customers make decisions. The only strategy that might work requires identifying what AI genuinely cannot replicate - and building everything around that. For some companies, it's taste and curation developed over decades. For others, it's relationships built on trust that can't be algorithmically reproduced. For still others, it's creative problem-solving that emerges from understanding context machines don't perceive. But you need to identify your irreplaceable advantage now. Because once you're dependent on AI platforms for customer discovery, you're competing on the terms AI optimizes for: price, availability, specs, and reviews. If that's your differentiation, you're a commodity. And commodities get squeezed. Three Questions To Consider Every board and C-suite needs to answer these questions this quarter: 1. When customers use AI to find what you offer, does your business get recommended - and how do you know? Most businesses have no visibility into AI recommendation algorithms. They can't optimize for selection criteria they can't see. They can't track whether they're winning or losing in conversations they're not part of. Start by actually testing: Ask ChatGPT, Gemini, and Claude for recommendations in your category. Does your business appear? In what context? With what framing? That's your baseline. If you're not being recommended, you don't exist to customers using AI. If you are being recommended, you have no control over how you're positioned. The second-order question: What makes you algorithmically discoverable? AI platforms optimize for factors you may not control - review volume, pricing transparency, data accessibility, integration ease. Your competitive advantages—service quality, expertise, relationships—may be invisible to recommendation algorithms. 2. What percentage of your customers will never visit your website because AI handles the entire interaction? Conservative estimates: 20-30% of customer research already happens through AI conversations. OpenAI projects 3 billion monthly active users by 2030. That’s not incremental. That’s a fundamental shift in how customers make decisions. Your acquisition funnel -website traffic, SEO rankings, conversion optimization - becomes irrelevant when customers never reach your site. The question isn't whether this happens. It's whether you've built any infrastructure for the post-website world. Can you be discovered, evaluated, and selected entirely through AI-mediated interactions? If not, you're optimizing for a channel that's already declining. 3. If AI platforms take 15% of every transaction they facilitate, do your economics still work? OpenAI needs alternative revenue streams beyond subscriptions to reach $125 billion by 2029. Transaction fees are the most obvious path. When AI completes a hotel booking, arranges professional services, or facilitates a purchase, why wouldn't the platform take a commission? The parallel: Expedia charges hotels 15-25% commission. Uber takes 25-30% from drivers. App stores claim 15-30% of digital transactions. AI platforms will charge for access to customers - probably starting lower, then increasing as dependency grows. Model your business with a 15% platform fee on every AI-originated transaction. Then model 20%. Then 25%. At what point do your margins collapse? That’s your leverage point for negotiation. Which is to say: you have little to none. When billions of transactions flow through a handful of platforms, you accept the terms or opt out. The Truth About "Everything Systems" AI platforms aren't building better tools. They're building the infrastructure that decides whether your business gets called at all. These platforms started as "chatbots" but are rapidly evolving into "everything systems" with unlimited memory, universal context windows, and reasoning capabilities that ensure users never need to leave the interface. Your business becomes a function that AI systems call. When someone asks for what you offer, the AI evaluates options and decides whether to invoke you. You don't control the evaluation criteria. You don't influence the recommendation logic. You don't even know you were considered and rejected. Integration feels like surrender. Resistance feels like extinction. There's no good answer. Just the one you can build a business around. Because the platform has already made its choice. Your customers are making theirs. The only question left: What are you willing to give up to stay in business when AI decides whether customers ever find you? Editorial StandardsReprints & Permissions
