Most AI Dev Tools Target Startups – the Money Is in Existing Ones

01 /The Concept

Prototyping new features for an existing product has a hidden constraint that general-purpose AI builders consistently fail to respect: the new pages have to look like they actually belong. Alloy's entire design philosophy starts from that constraint.

That's the fundamental difference between Alloy and tools like Lovable and other AI platforms that build apps and web services from scratch based on user prompts. Give Lovable a screenshot of your product as a reference and it'll use it as loose inspiration – then freely interpret and redesign as it sees fit.

Alloy takes a fundamentally different approach. It first analyzes the design system underlying the product and the component library from which its pages are built. Only then does it generate new pages with the requested functionality – pages that are visually indistinguishable from the existing product.

By default, Alloy doesn't even expose the generated code to users, and there's no way to export it from the platform. The output is exclusively "live pages" in the browser – where you can type into fields, click buttons, navigate between pages, and interact with everything exactly as if these were part of the real product.

But why would anyone need this?

The target audience is product managers. They can use Alloy to prototype new functionality for existing products – then share and discuss it with teammates or users.

The workflow:

- First, install a browser extension that analyzes existing product pages to build an internal model of the product's design system and UI component library.

- Then send natural-language requests to Alloy's AI assistant in chat – describing what new functionality should go where.

- If the result looks good, the product manager can share links to the new Alloy-generated pages with colleagues or users for review.

Viewing pages created by others is free with no restrictions. You only pay for users who can create pages – $20 per month per seat.

Alloy went through Y Combinator in winter 2023 and raised a new $2.2M round (approximately $3.5M AUD) in February. The startup published an announcement about the current platform's launch on the YC website a few days ago, claiming that the new version is gaining traction quickly among product managers at large companies and well-known startups.

02 /Why It Matters

Interestingly, right up until the recent launch announcement, today's Alloy was building an entirely different product – though also aimed at product managers.

The previous product was a platform for PMs to build work plans, prioritize tasks, and hand them off to designers and engineers – including specs for prototyping new features in existing products.

Then vibe coding platforms arrived and fundamentally changed what being a product manager means. PMs can now DO things, not just plan them. With vibe coding tools, they can design and build themselves rather than only delegating those tasks to others.

So Alloy rethought its entire concept – building a new platform to support the new workflows of a new generation of product managers who can actually create things with their own hands.

This shift has been widely discussed in the product community. Articles with titles like "Vibe Coding and the New Era of Product Management" and "Vibe Coding: The Hidden Superpower for Product Managers" have been multiplying online.

The change has touched every product discipline, and a number of startups building tools for product teams have been pivoting in this direction.

For example, Ion Design ([related review](/review/hochesh-sdelat-svoj-startap-novym-jepplom)) went through Y Combinator in winter 2024 and launched a new version of its product design platform last summer. Its AI first studies user flows through product logs, then proposes improvement options as ready-to-discuss prototypes for the team.

After review and sign-off, the same AI generates the code for the changes – which can be merged directly into the main codebase. And to make that merge painless, the AI first analyzes the existing code structure and libraries to ensure full compatibility.

So the real play for both Alloy and Ion Design is the same: don't create something new in a vacuum – create something new that evolves what already exists as smoothly and painlessly as possible.

That's actually what everyone working on existing products in startups and companies does every day. Yet it's something many developers of flashy new AI platforms for vibe coding, design, and other product tasks too often forget.

03 /Opportunities

Every business – large or small, a startup or a solo freelancer – isn't just a team of people completing tasks. It's a collection of established processes running on top of an established set of IT platforms and tools.

That's why any developer who builds a new business platform often can't sell it right out of the gate – it can't simply be plugged into existing processes and infrastructure.

This problem has come up repeatedly. A [related review](/review/bez-jetogo-uzhe-ne-prodat) covered WorkOS – the platform born out of the exact same frustration when the popular email client Nylas tried to sell to enterprise customers. Companies wouldn't buy it because they couldn't integrate it into their corporate IT infrastructure – for things like authenticating users through enterprise identity systems. So the founders built WorkOS, a platform that lets any web service become enterprise-ready with just a few lines of code. That idea attracted $95M in investment.

A more recent example is Swish ([covered here](/review/otlichnaja-fishka-v-interesnoj-nishe-na-rastushhem-rynke)), which built a sampling platform for e-commerce stores. It's been gaining traction precisely because it slots seamlessly into existing store workflows. The platform's AI:

- automatically intercepts each new order,

- decides which product to send as a sample based on the customer's purchase history – and only from items currently in sufficient inventory,

- adds that product to the order and passes it through for fulfillment as normal,

- then monitors subsequent orders from that customer to see if they start buying the sampled item – feeding that data back into sampling effectiveness reports and improving future selection algorithms.

Alloy and Ion Design belong in exactly this category. Their job is to create something that can be integrated as quickly and seamlessly as possible into what already exists – whether that's an existing software product, an enterprise IT infrastructure, or a business workflow.

The problem is that most developers of trendy AI business platforms build them as if for someone starting from scratch with nothing – as if their users can adopt the platform however the developer thinks best.

In reality, every such platform needs to be carefully woven into existing business processes, IT infrastructure, and existing products. That requires separate thinking and separate effort.

So the broader trend is clear: any new business product must "merely" improve what's already there – without forcing businesses to tear down and rebuild. Even if that new product is built on cutting-edge AI.

The most durable entry angle: pick a process that businesses run every day, map the exact friction points where new tools get rejected because they don't integrate, and design for insertion rather than replacement. The platform that makes itself invisible – fitting into what already exists rather than demanding adaptation – is the one that actually gets adopted.