ChatGPT Becomes an App Platform as OpenAI Challenges Mobile Distribution

The Interface Becomes the Store

OpenAI's developer conference this week introduced autonomous agents called Dots and updated AI models, but the more structural shift lies in how the company is repositioning ChatGPT itself. The chatbot, which OpenAI says now reaches 1.2 billion people each week, is evolving into a surface where software can be found, authenticated and executed without users ever opening a traditional app.

The mechanics work like this: ChatGPT will suggest applications mid-conversation when it detects that a specific tool could help finish a task. Users can connect the app immediately and operate it within the chat window. OpenAI has expanded its plugin architecture to support what it calls extensions, which let developers build interactive panels that sit alongside the conversation thread. Tasks that previously required switching to a browser tab or launching a mobile app now happen inside ChatGPT's interface.

At Opentechwire, we've tracked similar moves in Asia where super-apps like WeChat and Grab collapsed multiple services into single platforms, but those efforts began with payments or ride-hailing and expanded outward. OpenAI is attempting the inverse: starting with a conversational interface and layering commerce, productivity and design tools on top.

Identity and Allowance Portability

The company introduced "Sign in with ChatGPT," an authentication system that carries a user's existing AI usage allowance into third-party applications. Sixteen partners joined at launch, including Cognition's Devin code assistant, Notion, Vercel, and design collaboration tools. The allowance portability matters because it removes a friction point: users no longer need separate subscriptions or credits for each AI-enabled service they access through ChatGPT.

OpenAI demonstrated a meeting scheduler built into ChatGPT that pulls calendar data, offers to transcribe calls using AI, and then surfaces action items after the session ends. Another demo showed design workflows with Figma and Adobe integrations, where revisions and feature requests happen without leaving the chat. These lightweight apps can be shared as ChatGPT-generated websites, letting colleagues sign in with their own credentials and permissions.

The identity layer is critical infrastructure. It positions ChatGPT not just as a front-end for isolated tasks but as a persistent environment where work happens and permissions follow the user across tools.

Developer Submission and Marketplace

OpenAI has refined its plugin review process, adding status tracking, itemised feedback, and the ability to update tools without resubmitting the entire package. The company also launched an enterprise app marketplace with more than 30 partners at debut, including Adobe, Figma, Salesforce, HubSpot, ServiceNow, and security vendors like Palo Alto Networks and CrowdStrike.

Eligible enterprise customers can apply portions of their OpenAI spending commitments toward approved partner software, creating a bundled procurement channel that mirrors enterprise app stores but runs through OpenAI's billing relationship.

For developers, the calculation is straightforward: as more discovery shifts into conversational interfaces, building an AI-native version of their product for ChatGPT becomes a distribution imperative. The chatbot's weekly user base is larger than most operating systems, and the interface reduces the steps between intent and action.

Agents as App Selectors

The Dots autonomous agents add another dimension. Each Dot operates in its own cloud environment with a browser and can connect to any of the more than 4,000 apps now integrated with OpenAI's ecosystem. Users describe a goal, and the agent selects the necessary tools, models or services to complete it.

This flips the traditional app discovery model. Instead of users searching an app store, downloading software, and configuring it, the agent presents the tool only when needed. OpenAI says Dots will perform background research in read-only mode across connected apps, then surface proactive suggestions - always requiring user approval before taking action.

The implication for app makers is that their software may be invoked by an agent rather than deliberately chosen by a person. Subscription and billing still happen outside OpenAI's system; users would subscribe via the app maker's website and then connect that account to ChatGPT or the agent. OpenAI has not announced any revenue-sharing or payment processing layer comparable to Apple's or Google's 15 to 30 per cent cut.

What's Missing: The Economic Layer

OpenAI's developer announcements focused on discovery, distribution, identity and execution - four pillars of a platform. The fifth pillar, monetisation infrastructure, remains undefined. The company has not introduced a billing system that would let developers sell subscriptions or in-app purchases directly through ChatGPT, nor has it outlined a revenue split for transactions that originate in the chat interface.

That omission is deliberate or incomplete. If OpenAI eventually layers in payment processing and takes a share of transactions, it would replicate the full app store stack and compete directly with Apple and Google on economic terms. For now, the company is building user behaviour and developer participation without claiming a toll.

The enterprise marketplace offers a partial model: bundled procurement where OpenAI credits can be spent on partner software. Extending that to consumer transactions would require fraud controls, refund policies, and regional payment compliance - problems Apple and Google spent years solving.

Distribution Power in Conversational Interfaces

The structural advantage OpenAI is pursuing is attention. Users already spend time in ChatGPT for research, drafting and problem-solving. If that interface can also handle scheduling, design iteration, code generation and data analysis without context-switching, the chatbot becomes the default workspace.

App stores succeeded because they centralised discovery and reduced friction. OpenAI is arguing that conversational AI reduces friction further: users describe intent in natural language, and the system assembles the necessary tools in real time. The app becomes invisible; the outcome is what matters.

This model works best for tasks that are well-defined and can be automated or assisted by AI. It works less well for exploratory software use, entertainment, or experiences where the interface itself is the product. A meditation app or a game is not easily collapsed into a chat thread.

But for productivity, data manipulation, and workflow automation - categories that dominate enterprise software spending - the ChatGPT platform model is coherent. Developers gain access to a massive user base with built-in AI capabilities. Users gain a single environment where tools appear on demand and credentials travel with them.

Regional Context and Competitive Response

In Seoul, Singapore and Shenzhen, platform consolidation is a familiar pattern. Naver, Grab, and Tencent built ecosystems by embedding services into high-frequency interfaces. OpenAI's approach is similar in structure but different in entry point: instead of payments or transport, the anchor is conversational AI.

Google has the most to lose in the near term. Its search business already faces pressure from AI-driven answer engines, and Android's app distribution model could be undermined if users begin interacting with software primarily through chat. Apple's iOS ecosystem is more vertically integrated, but even there, the shift from app icons to agent-mediated tasks would erode App Store centrality.

Microsoft, an OpenAI investor, benefits indirectly. Its enterprise software portfolio, including GitHub, Office and Azure, can integrate with ChatGPT's platform features, and the company has already embedded OpenAI models across its product line.

The question for developers is whether to build for this new surface or wait. The answer likely depends on how much of their user acquisition currently happens through search or conversational queries. If a significant share of traffic already comes from people asking questions rather than browsing, an AI-native version of the product is worth the investment.

What Comes Next

OpenAI has not indicated when it might introduce payment processing or revenue-sharing for apps distributed through ChatGPT. The company also has not addressed how it will handle app quality, fraud, or disputes - governance challenges that scale with platform size.

The enterprise marketplace offers a testing ground. If bundled procurement proves popular and OpenAI can manage partner relationships at scale, a consumer-facing version becomes more plausible. The identity layer and allowance portability are already in place; adding payments would complete the stack.

For now, the company is focused on building developer participation and user habits. The Dots agents, the plugin extensions, and the Sign in with ChatGPT feature are all infrastructure that makes the chatbot stickier and more capable. Whether that infrastructure eventually supports a full economic platform or remains a distribution channel for third-party billing is still open.

What is clear is that OpenAI is not content to remain a model provider. The company is building the layers - discovery, identity, execution, and eventually monetisation - that would let it operate as a platform in its own right, independent of mobile operating systems and their app stores.