Edinson Carranza

SaaS Isn't Dead. But Your Business Model Might Be. What Should We Build in 2027? For more than a decade, building a SaaS company was one of the most attractive business models on the internet. You developed an application. You acquired users. You charged $19, $49, or $99 per month. And if you managed to retain your customers, you could build a business with recurring revenue, predictable cash flow, and potentially extraordinary margins. The formula was relatively simple. Acquire customers. Increase MRR. Reduce churn. Scale. Thousands of entrepreneurs built successful companies following this model. But I believe we're entering a period where the economics of software are about to change dramatically. Not because subscriptions are disappearing. But because artificial intelligence is fundamentally changing what customers are willing to pay for. And this raises a question every developer, indie hacker, and startup founder should be asking: What should we actually build in 2027 if AI can increasingly replicate the software we're trying to sell? Let's dig deeper. 1. When Building Software Becomes Easier, Selling Features Becomes Harder For years, one of the biggest competitive advantages of a technology startup was its engineering capability. Building a CRM, analytics platform, automation tool, or enterprise application required developers, designers, infrastructure, and months of development. Today, AI coding agents are dramatically accelerating that process. And there's an interesting signal emerging from the market. According to McKinsey's 2026 research on artificial intelligence, 32% of respondents reported that their organizations had decided not to purchase at least one software product or feature because they could build it internally using coding agents. Think about what that means. For decades, companies purchased software because developing it internally was too expensive, too complicated, or too time-consuming. Now, some organizations are beginning to reconsider that decision. Why pay a recurring subscription for a relatively simple feature if your engineering team can build a customized alternative? Of course, generating code isn't the same as maintaining reliable software. Security, compliance, integrations, testing, and long-term maintenance still matter. But the economic equation is changing. And when the cost of creating software decreases, the value of certain standalone features can decrease with it. The problem isn't that software is becoming worthless. The problem is that software features alone are becoming less defensible. 2. The Next SaaS Crisis Could Be a Crisis of Differentiation Open Product Hunt. Every week, hundreds of new products compete for attention. AI writing assistants. Content generators. Analytics dashboards. Marketing automation tools. AI-powered CRMs. Productivity applications. Research assistants. Many of these products are genuinely useful. But there's an uncomfortable question founders need to ask themselves. What happens when the AI model powering your application introduces the same functionality directly? What happens when your competitors can reproduce your core features in a weekend? What happens when customers discover they can accomplish similar tasks using tools they're already paying for? Where does your competitive advantage go? I'm not suggesting every AI wrapper will fail. Some wrappers become valuable companies because they offer exceptional user experiences, specialized workflows, distribution, integrations, or customer relationships. But simply connecting an API to a beautiful interface is rarely enough to create a durable business. And I believe this will become increasingly obvious in 2027. The market won't reward every application just because it uses artificial intelligence. It will reward companies that deliver value customers cannot easily obtain elsewhere. 3. We're Moving From Software as a Service to Outcomes as a Service This is where things become interesting. Traditional SaaS primarily sells access to software. You pay for seats. You pay for features. You pay for storage. You pay for premium functionality. But AI agents introduce a different economic model. What if customers paid for completed work instead? Imagine a company paying $500 per month for a customer support platform. The software helps employees manage conversations. Now imagine an AI system that autonomously resolves hundreds of customer requests. Should the company pay for access to another dashboard? Or should it pay for successfully resolved requests? Consider accounting. Instead of selling a tool that helps accountants process invoices, imagine selling a system that processes invoices, validates information, detects errors, and updates financial records. The customer isn't necessarily interested in the software itself. The customer wants the work completed. This is the opportunity. The next generation of software companies may increasingly sell completed work rather than access to tools. But there's an important distinction. Not every outcome is easy to measure. Not every task can be automated reliably. And not every business can afford the unpredictable costs of autonomous agents. That's why pricing models will need to evolve alongside the technology. 4. What Products Should Developers Build for 2027? I believe several categories deserve serious attention. Vertical AI Agents Not another generic chatbot. Not another AI assistant that promises to do everything. I'm talking about specialized agents designed for specific industries. Healthcare administration. Logistics. Accounting. Insurance. Real estate operations. Legal workflows. Financial operations. These industries have complex processes, specialized terminology, regulatory requirements, and expensive operational inefficiencies. A product that understands a specific industry's workflows and reliably executes valuable tasks could become significantly more defensible than a generic productivity application. The opportunity isn't necessarily to replace entire teams. It's to automate repetitive work while preserving human oversight where judgment, accountability, and expertise matter. Autonomous Business Operations Imagine software that doesn't simply display information. It acts on it. A system receives a document. Extracts the relevant data. Validates the information. Identifies inconsistencies. Updates internal systems. Requests approval when necessary. And completes the workflow. No constant switching between dashboards. No repetitive manual data entry. No need for employees to coordinate every individual step. This is a different way of thinking about enterprise software. Instead of building another application employees must learn to operate, we're building systems that perform portions of the work themselves. Infrastructure for AI Agents Here's an opportunity that I think many developers are underestimating. If companies begin deploying dozens or hundreds of AI agents, who will manage them? Who controls their permissions? Who monitors their behavior? Who tracks their costs? Who verifies their outputs? Who prevents agents from accessing sensitive information? Who determines which actions require human approval? As autonomous systems become more capable, businesses will need infrastructure to manage reliability, security, observability, governance, and spending. Some of the most valuable opportunities may come from solving problems created by AI adoption itself. Revenue Automation Companies don't necessarily want more marketing tools. They want more revenue. They want qualified leads. They want fewer cancellations. They want higher customer retention. They want unpaid invoices recovered. They want sales processes that operate more efficiently. This creates opportunities for products that connect AI capabilities directly to measurable business outcomes. But the key is attribution. If a product claims to increase revenue, it needs a credible way to demonstrate that contribution. Otherwise, outcome-based pricing becomes another marketing promise. AI Cost Optimization There's another emerging challenge. AI makes software easier to build, but operating AI-powered products introduces variable costs. Inference. Tokens. Compute. Model routing. Storage. Monitoring. Retries. Agent execution. A company can generate impressive revenue while quietly destroying its margins through inefficient AI usage. Products that help businesses optimize model selection, manage inference spending, and improve cost efficiency could become increasingly valuable. Because in the AI economy, revenue growth without cost discipline can become dangerous. 5. How Should We Monetize These Products? This is where I think founders need to rethink the traditional SaaS playbook. For years, we focused almost exclusively on monthly subscriptions. And subscriptions remain powerful. Predictable recurring revenue is valuable. But charging a fixed monthly fee for unlimited AI usage can become a serious financial risk. Every agent execution has a cost. Every API call consumes resources. Every automated workflow may involve multiple models, external services, and infrastructure. The more successful your product becomes, the more expensive it may become to operate. This creates a very different economic reality from traditional SaaS products with relatively low marginal usage costs. So what should we do? I believe hybrid pricing models will become increasingly important. A monthly platform subscription combined with usage-based billing. Enterprise contracts with defined execution limits. Pricing based on successfully completed tasks. Consumption credits. Or performance-based fees when results can be independently verified. For example: An invoice automation platform could charge $199 per month plus a fee for every successfully processed document. A sales automation product could charge a monthly platform fee plus an agreed amount for qualified opportunities. An enterprise agent platform could charge for execution capacity, monitoring, and security features. The important thing isn't choosing the trendiest pricing model. It's aligning the price with the value delivered while protecting gross margins. Because generating $100,000 in monthly revenue means very little if delivering the service costs $95,000. And a subscription model that looks attractive on a Stripe dashboard can still be economically unsustainable. 6. The Future of MRR Might Look Different Let's imagine two startups. Startup A sells a generic AI productivity tool for $15 per month. It needs 2,000 paying customers to generate $30,000 MRR. Startup B sells a specialized AI workflow system to businesses. It charges $200 per month plus an average of $100 in monthly usage. It needs 100 active customers to generate the same $30,000 MRR. Both models can work. Neither is automatically superior. Startup B may face longer sales cycles, more complex onboarding, higher support requirements, and greater customer concentration risk. Startup A may have lower acquisition friction and a much larger potential audience. But the comparison illustrates something important. You don't necessarily need thousands of subscribers to build a meaningful recurring-revenue business. You need customers who receive enough value to justify what they pay. And increasingly, that value may come from work completed rather than features unlocked. 7. My Prediction for 2027 I don't believe SaaS will disappear. I believe its value proposition will evolve. From tools to specialized operational systems. From features to workflows. From active users to completed tasks. From exclusively fixed subscriptions to hybrid monetization. From generic AI applications to products deeply integrated into specific industries. And from selling software access to delivering measurable business value. But there's something even more important. Artificial intelligence is dramatically increasing the supply of software. It is not necessarily increasing demand at the same rate. We can build ten times more applications without creating ten times more problems worth solving. And that's the fundamental challenge facing the next generation of technology entrepreneurs. The ability to build is becoming increasingly accessible. The ability to identify valuable problems, acquire customers, operate efficiently, and maintain competitive advantages remains difficult. That's why I believe the most successful founders of 2027 won't necessarily be those launching the most products. They'll be the ones who understand their customers better than everyone else. Who build reliable systems. Who solve expensive problems. Who create measurable economic value. And who design business models capable of surviving technological change. 8. The Question Every Founder Should Ask Before Building Another SaaS Before writing another line of code, ask yourself: What recurring problem am I solving? How much does that problem cost the customer? Can artificial intelligence solve it reliably? How will I measure the results? Why would customers choose my product over a generic AI agent? What prevents competitors from replacing me? Can I deliver the service profitably at scale? And most importantly: Will customers still have a reason to pay for my product when AI models become ten times more capable and significantly cheaper? Because that moment may arrive faster than many founders expect. And when it does, simply having an AI-powered product won't be a competitive advantage. It will be the baseline. The real opportunity won't be building another subscription-based application. It will be creating systems that customers depend on because they consistently solve important problems. We shouldn't stop building SaaS. We should stop assuming that recurring billing automatically creates recurring value. The next generation of technology companies won't be defined only by how many subscriptions they sell. They'll be defined by how much useful work they can deliver, how reliably they deliver it, and whether they can do so profitably. SaaS isn't dead. But the era of charging recurring fees for easily replaceable features is under pressure. And that might be one of the biggest opportunities developers and founders have ever had. Keep building. But build something people will have a reason to keep paying for.

The Binance hack is totally an inside job... How do you guess the right authorization number from limitless possibilities? Clearly they just sprayed 50 to build the narrative that "the attacker just got lucky..."