# The Dodo Digest: The AI Bubble Isn't What You Think > AI feels like a bubble, but revenue is finally catching up to the hype. The real risk isn't AI - it's how companies choose to use it. We break down why short-term optimization creates fragile systems, and what we've shipped to help builders work with AI inside real systems. - **Author**: Rishabh Goel - **Published**: 2026-05-06 - **Category**: Newsletter - **URL**: https://dodopayments.com/blogs/newsletter-may6 --- **TL;DR** - AI feels like a bubble, but revenue is finally catching up to the hype. - At the same time, products are getting replaced faster than ever, and companies are reacting short-term. - Some are cutting jobs aggressively, while others (like China) are pushing back. - The real risk isn't AI, it's how companies choose to use it. - We've been building for a different future: one where AI works with systems, not just replaces them. **Hello everyone,** I was reading an article from The Atlantic earlier this week about whether the AI bubble is about to pop. Few months ago, the concern was overinvestment. Too much money going into infrastructure without clear returns. Now it's the opposite. AI tools are being adopted so fast that companies can't keep up with demand, and revenues are growing at a pace that's honestly hard to process. And yet, at the same time, something else feels off. Every time a new "breakthrough" product drops, it gets replaced just as quickly. And that's when it clicked. This isn't a normal cycle. ## The Signal We're in a strange phase of the AI cycle. On one side, the numbers look very real. Adoption is accelerating faster than expected, and companies are seeing tangible productivity gains. Tools like Claude Code have pushed AI beyond assistance into execution, where tasks that used to take days can now be completed in hours. But on the other side, the pace feels unstable. Products don't last. Advantages don't hold. What feels like a breakthrough one week quickly becomes the baseline the next. The cycle from innovation to commoditization has compressed dramatically. That creates a different kind of pressure for companies. When everything moves this fast, it becomes difficult to build with a long-term view. Decisions start getting driven by what works immediately, what improves metrics now, what reduces costs this quarter. And that's where things start to break. Because optimizing for speed without stability creates systems that look efficient on the surface, but don't hold up over time. ## Where Things Start to Break We're already seeing early signs of this. Some companies are starting to replace employees with AI not because the system is fully ready, but because it improves short-term efficiency. Others are restructuring teams around what AI can do today, without fully understanding what it can't do yet. At the same time, not every region is accepting this shift the same way. China, for example, has already pushed back legally against companies replacing workers purely with AI. That contrast matters. Because it shows this isn't just a technology shift. It's a decision about how we choose to build companies. And whether we optimize for short-term gains, or long-term systems. ## How To Think About This Here's how to approach this if you're building in AI right now: 1. **Don't mistake speed for stability.** Just because something works today doesn't mean it will hold tomorrow. AI capabilities are improving too quickly for assumptions to stay valid. 2. **Separate tasks from roles.** AI can replace tasks, not entire systems. Designing your company around full replacement often creates gaps you only see later. 3. **Avoid short-term optimization traps.** Cutting costs using AI might look efficient now, but it can damage product quality, trust, and long-term growth. 4. **Build systems people can trust.** Trust compounds over time. If users or teams feel instability, it affects adoption more than any feature improvement. 5. **Think in decades, not cycles.** The companies that win won't be the fastest to react, they'll be the ones that build systems that last. ## What This Means for Builders This is where the narrative gets more balanced. While a lot of the conversation focuses on AI replacing jobs, there's another perspective emerging. Jensen Huang recently pointed out that AI is also creating entirely new categories of work. Not just around models, but around infrastructure, systems, and everything that supports them. And that makes sense. Because every time a new layer of technology is introduced, it doesn't just remove work. It reshapes it. The real question isn't whether AI replaces jobs. It's whether companies build in a way that expands opportunity, or compresses it. ## What We Shipped This is something we've been thinking about deeply. If AI is going to become part of how systems operate, then it shouldn't just replace workflows. It should be able to work within them. That's exactly why we built our [MCP server](https://docs.dodopayments.com/developer-resources/mcp-server#mcp-server) and [agent plugin](https://docs.dodopayments.com/developer-resources/mcp-server#agent-plugin-recommended). Instead of exposing dozens of disconnected APIs, it allows AI agents to interact with your payment system in a more structured way. Agents can fetch context, execute actions, and handle multi-step workflows directly through code, rather than requiring constant back-and-forth. What's changed recently is how real this has become. There was a time when agentic workflows felt experimental. Today, they're driving real usage at scale. Tools like Claude have already crossed $30 billion in annualized revenue, which says a lot about how quickly this shift from "chat" to "action" is happening. We're seeing the same shift with builders. Startups using Dodo are already integrating agents into their workflows, not as a layer on top, but as part of how their systems operate. Payments, subscriptions, customer actions, all handled through agents with minimal friction. A big part of making this work is simplicity. Our [docs](https://docs.dodopayments.com/introduction) are structured so that agents can understand them directly. In most cases, all you need to do is pass the documentation link, and the agent has enough context to start interacting with the system, no complex setup, no manual mapping of endpoints. The idea is simple. AI shouldn't just generate output. It should be able to operate reliably inside real systems. ## One Last Thought The AI bubble question misses something important. This isn't just about whether the market is overvalued or not. It's about how we choose to build within it. Because if everything is optimized for short-term gains, the system becomes fragile. But if it's built on trust, stability, and long-term thinking, it compounds. And in the end, that's what lasts. Also, join our loving [Discord community](https://discord.gg/dodo-payments-1305511580854779984)! Best, Rishabh Goel Co-Founder, **Dodo Payments** --- - [More Newsletter articles](https://dodopayments.com/blogs/category/newsletter) - [All articles](https://dodopayments.com/blogs)