# Product is dead. Long live the new product. > Question: What kind of product should I build in the AI era, now that web apps are dead? > Author: Adi Shmorak, The P/MF Detective (https://adidacta.com/about) > Updated: 2026-10-29 > Canonical: https://adidacta.com/questions/product-is-dead-pipeline-or-harness **Short answer:** The web app is no longer needed to deliver the value. In the AI era, the value comes from the AI machinery behind it, and most new products take one of two shapes: a pipeline, which delivers an outcome, or a harness, which gives a team control over a focused process. Which one you build depends on who uses it, and that choice shapes your ICP, value proposition, roadmap and business model. ## Related questions - [The P/MF Matrix: 4 steps, 3 tools, 12 questions to find product/market fit](https://adidacta.com/questions/pmf-matrix) - [How do you find product/market fit for an AI product?](https://adidacta.com/questions/pmf-for-ai-products) - [What should a product team build before product/market fit?](https://adidacta.com/questions/what-to-build-before-pmf) [← All case files](https://adidacta.com/questions) --- ## Is the web app really dead? Everybody says SaaS is dead. But SaaS is a business model, not a product. What's really fading is the web app as the machinery that creates the value. And that's exactly why SaaS looks like it's dying. On [the road to P/MF](https://adidacta.com/questions/pmf-matrix), a change in one step cascades to every step after it: change the product, and the business model has to change too. A web app cost almost nothing to run per user, so a flat monthly subscription worked. AI costs money every time it works, so a flat, unlimited price would bankrupt you. That's why AI products cap usage or charge per use. It's not that SaaS was wrong. The product changed, and the business model has to follow. Think of a product as an elevator. Your users walk in, press the button for the floor they want, say floor 37, and expect to get exactly there. No hassle, no friction. Delivered, done. Like a pill that takes away a headache. That's why the elevator is my perfect analogy for a product. It's the machinery that delivers the transformation, and the transformation is your value proposition. Your users don't care how the sausage gets made. They care about reaching floor 37 with as little effort as possible. Web apps were the stairs. They gave users a nice UI, but the users still did most of the climbing. Think of designing in Figma: getting from a blank canvas to a finished design takes a lot of effort. With Claude Design, it's a few prompts, some inspiration and a bit of guidance. The AI does the climbing. That's why the web app, as the machinery that creates the value, is dead. ## One decision sets everything else So what replaces the web app? AI machinery that does the climbing for your users. It comes in a few shapes, and choosing yours is the one decision that sets most of the others for you: who you sell to, what you promise, what you build first and how you charge. Most founders make it without noticing, and pay for it later with a muddled roadmap and pricing that doesn't fit. Make it on purpose and everything after it gets easier. And it's a short list: in the AI era, two new options replace the web app. ### Option 1: the pipeline > **A pipeline sells an outcome.** You give it an input, it runs a fixed recipe on its own, and it delivers the same kind of result every time. A pipeline is a fixed recipe. You know the input, the output, and the steps in between. It runs on its own, only asks you for input when it needs you, and delivers the same kind of outcome every time. You hand over control of the process and trust the recipe. Here's one I built. You give it your company's domain. It researches your ICP from your website, studies your customers and competitors, builds lookalikes, turns them into filtering criteria, finds prospects, qualifies them, and identifies the buyer persona. Out comes a list of 50 ICP-fit companies with the right person at each. Every week, or every day, depending on your budget. I built it for a client as a lead magnet. The same pipeline could just as well be a paid product. Product type and business model are separate decisions. Because it sells an outcome, a pipeline can be priced on outcomes: for example, a monthly price for up to a set number of qualified leads. ### Option 2: the harness > **A harness sells control.** A focused workspace where you steer the process, with the tools, rules and context to do many related jobs and run many pipelines. A harness is looser. It's a workspace, often a chat, where you control the process and can do many related things within clear limits. It holds the tools, the rules and the context, and it can run many pipelines. ChatGPT is a general harness: anything goes. The go-to-market harness I'm building for RallyUp is the opposite. It's built for LinkedIn outreach only. From one workspace you launch campaigns, enrich data, and run connection and DM campaigns. It knows its tools, it watches costs, and it can tell a good ICP fit from a bad one. That focus is what makes the control useful. Because it sells control, a harness is usually priced on usage: a fixed plan with limits, or credits, the way Claude Code works. ## So is the web app truly dead? As the machinery that creates the value, yes. As a screen, no. It stays, with a new job. AI machinery is a black box. Users still want to see inside it, and a chat isn't always enough. So the web app becomes the viewport: a control panel where you see context, progress and results at a glance, turn the knobs that tweak what the AI does, and get visibility into what's going on inside its process. For the seller, that screen matters more than it looks, because it shapes perceived value. Visible work feels like more value than work that happens out of sight. And perceived value sets a lot: how much customers are willing to pay, how long they stay, and so your churn and customer lifetime value. Higher perceived value means higher willingness to pay. Always. So the web app no longer creates the value, but it still decides how much of that value your customers feel, and pay for. ## How to choose: are you serving the diner or the chef? Picture a restaurant. A pipeline is a dish on the menu: order it and you get the same dish every time. A harness is the kitchen: the ingredients, the tools and the rules, where you can make many dishes, including new ones. The diner wants a great dish. They don't care how it's made. They want something tasty on the plate. The chef is different. The chef wants to create new dishes and control every ingredient and every step. If your user is the diner, build a pipeline. "I need 200 ICP-fit leads a month to feed my SDRs." They want the result, not the process. If your user is the chef, build a harness. An agency with many clients wants to experiment, change the inputs, and run different plays for each client. It's the user who decides, not whoever signs. A CEO who never cooks might still buy a harness, because the team in the kitchen wants control. ## A third path Some founders won't build either. They'll sell a productised service, with AI doing much of the work behind a person. That's a different article. ## Your checklist 1. Write down the transformation your product delivers: which floor does your customer want to reach? 2. Ask who uses it: a diner who wants the outcome, or a chef who wants control of the recipe? 3. Diner: build a pipeline and price on outcomes. Chef: build a harness and price on usage. 4. Design the UI to show the value, not to do the work. --- About the author: Adi Shmorak is a fractional CPO and Product/Market Fit consultant (https://adidacta.com/about). Adi has worked with over 100 founders since 2019. He has 74 published testimonials from founders, accelerator leads and product executives (GrowthMentor, LinkedIn, Google, Facebook and direct), and a 5.0/5 average across all 47 reviews on GrowthMentor (https://www.growthmentor.com/mentors/adi-shmorak). Full text: https://adidacta.com/testimonials.md