Konstantin Semenenko
July 14, 2026
4
minutes read
With Apps in ChatGPT, products now run inside the conversation, a user can book, buy, design, or check something without leaving the chat. For SaaS, this is a new distribution and interface surface reaching hundreds of millions of users, and it changes a core assumption: your UI is no longer the only way in. The strategic question is not "should we build a ChatGPT app" but "which one customer workflow should work inside a conversation first." The winners will expose a few high-value actions cleanly, keep their product as the source of truth, and build on the open standard (MCP) so the work is portable, not bet on one host.
Apps in ChatGPT mean that software products now run inside the conversation itself, and for SaaS that is a genuine shift, not a gimmick. A user talking to ChatGPT about a trip can book through an in-chat app; someone outlining a deck can turn it into slides without leaving; a customer with a support question can act on their account in the thread. With ChatGPT reaching hundreds of millions of weekly users, this creates a new interface and distribution surface where the conversation, not your dashboard, is increasingly where a task starts. The strategic implication for SaaS is concrete: your product's UI is no longer the only way customers reach its value, and the question every SaaS team should be asking is which single workflow should work inside a conversation first. This is what the shift actually means and how to respond to it.
We build the layer that puts product actions into AI conversations, our AI Product Interface work, so this is a practitioner's read on what Apps in ChatGPT change for SaaS and what to do about it.
The change is that the conversation became a place to act, not just to ask. Apps in ChatGPT, built on the Model Context Protocol, let a product expose interactive experiences and real actions inside the chat, so a user can complete a task, book, buy, design, check, order, without switching to a separate app. OpenAI shipped this with launch partners like Booking, Canva, Spotify, and Zillow, opened it to all developers through the Apps SDK, and has been building out submission, discovery, and distribution around it.
For SaaS, three things follow. First, there is a new surface where customers encounter your product, inside a conversation, suggested when relevant, not only on your website. Second, the interface model changes: instead of navigating your UI, a customer describes what they want and your product's actions run in the chat. Third, it is built on an open standard, so the same work can, in principle, run across hosts rather than being locked to one. The conversation becoming a runtime for products is the shift, and it reshuffles assumptions SaaS has relied on for a decade.
SaaS has always assumed the product's UI is the front door: customers log in, navigate, and act inside your interface. Apps in ChatGPT loosen that assumption. When a customer can accomplish a task by describing it in a conversation, the value of your product, the action it performs, the data it holds, becomes reachable without your UI being the path. That is both an opportunity and a risk, and which one it is depends on whether your product can take part in the conversation.
The opportunity is reach and reduced friction: a task that took five tabs (CRM, orders, calendar, support) becomes one request, and your product meets customers where they increasingly start. The risk is the mirror image: if a competitor's product is present in the conversation and yours is not, the conversation routes around you. For SaaS, "our product is not in the chat" is starting to look like "our product is not on mobile" did fifteen years ago, survivable for now, a growing disadvantage over time. The shift rewards products that show up where the work starts.
The wrong response is to bolt a chatbot onto your product and call it done. Apps in ChatGPT are not a chat window on your marketing site, they are your product's real actions running inside the AI host, with the host's authentication, UI, and approval, and your product's authorization and data behind them. Treating this as "add a chatbot" produces something that talks about your product instead of doing your product's work, which misses the entire point of the shift.
The better mental model, and the harder work, is exposing a small set of your product's real actions safely into the conversation: named tools, validated inputs, visible approval for consequential steps, your existing authorization preserved, and an inspectable record returned. That is product and platform engineering, not a widget, which is why we treat it as a distinct surface in what is an AI Product Interface. The teams that win here build the interface with the same rigor they build the product, because it is the product, reached a new way.
The move is not to rush a broad integration, it is to pick well and start narrow. The practical response:
Start with one workflow done well, and expand from evidence. A single high-value action working cleanly in the conversation teaches you more than a broad, shallow integration, and it is shippable now.
Apps in ChatGPT mean products now run inside the conversation, and for SaaS that is a new interface and distribution surface reaching hundreds of millions of users, one that loosens the old assumption that your UI is the only way in. The opportunity is reach and reduced friction; the risk is that the conversation routes around products that are not present in it. The wrong response is bolting on a chatbot; the right one is exposing a few of your product's real actions safely, keeping your product the source of truth, and building on the open standard so the work is portable. The question is not whether to engage, but which one workflow should work inside a conversation first.
If you want to put one high-value workflow into ChatGPT, Claude, and Gemini, on top of your existing SaaS product and under your own controls, that is exactly what our AI Product Interface work delivers, and the insights go deeper on the product choices behind it.
What are Apps in ChatGPT? Interactive product experiences that run inside a ChatGPT conversation, built on the Model Context Protocol, so a user can complete real tasks (book, buy, design, check) without leaving the chat. OpenAI launched them with partners like Booking, Canva, Spotify, and Zillow, and opened the Apps SDK to all developers.
What do Apps in ChatGPT mean for SaaS? They create a new interface and distribution surface where customers can reach your product's value inside a conversation, not only through your UI. That is an opportunity (reach, less friction) and a risk (the conversation can route around products that are not present), so being in the chat is becoming a competitive factor.
Should my SaaS build a ChatGPT app? The better question is which one customer workflow should work inside a conversation first. Rather than a broad integration, pick a single high-value task, expose your product's real actions for it safely, and expand from evidence. Start narrow, keep your product the source of truth, and build on the open standard.
Is a ChatGPT app just a chatbot for my product? No. It exposes your product's real actions inside the AI host, with named tools, validated inputs, visible approval, and your existing authorization, not a chat window that talks about your product. Treating it as "add a chatbot" misses the point; it is your product's actual capabilities reached a new way.
How do I make a ChatGPT app portable across AI hosts? Build the core on MCP, the open standard behind Apps in ChatGPT, and treat each host's authentication, UI, approval, and release rules as a separate layer. The hosts differ and their rules evolve, so building on the standard and layering host-specific behavior keeps the work portable rather than locked to one.
