Several announcements from OpenAI’s DevDay 2026 could become the basis of entirely new products. Others could give an established web or mobile app a useful next feature. We can help build with all of them.
The event on 29 September introduced more ways to build applications, reach customers through ChatGPT and pay for the AI those applications use. For founders and product teams, the question is where those capabilities solve a customer problem well enough to support a business.
These are the five developments we’d pay particular attention to. OpenAI’s DevDay 2026 recap covers the full set of announcements; here we’re focusing on what they could mean for products. Availability and access conditions below reflect the launch information as of 30 September 2026.
1. ChatGPT plugins: a new route to customers
OpenAI’s plugin extensions let developers add interactive panels, sidebar experiences and file viewers to ChatGPT. Improvements to discovery also aim to make relevant plugins easier to find.
The opportunity is to make your product available at the point someone needs it. A specialist service could be discovered while a prospective customer is asking for help with the problem it solves. An established platform could give existing customers a more convenient way to get things done.
That makes a useful ChatGPT integration worth considering as part of a product’s route to market. It also raises a familiar design question: which task would someone actually want to complete there? Moving an entire application into a conversation is less compelling than making one valuable task easy.
We explore the wider shift towards making existing web apps usable by AI agents in a recent insight. The product still owns its data, permissions and business rules, even when customers use it through another interface.
2. Agents API: more ambitious products, less infrastructure to build
The Agents API provides a managed foundation for agents that use tools and work through tasks. DevDay adds computer use and further multi-agent capabilities.
That could change the scope and cost of building a specialist AI product. More of the investment can go into understanding your customers’ workflow, implementing your business rules and designing the experience that makes the product worth using.
For a new business, this could help turn a manually delivered service into a repeatable software product. For an existing platform, it could support a premium feature that prepares a substantial piece of work for a customer to review.
Managed infrastructure does not remove the need for product engineering. An agent needs appropriate access, a clear stopping point and a way to recover when something goes wrong. Our Three Cs of Production AI explain how context, constraints and control shape that work.
3. Sign in with ChatGPT: a different way to fund AI features
Sign in with ChatGPT lets eligible users authorise participating tools to draw on their ChatGPT plan allowance. Identity and permission to use that allowance are separate capabilities, and usage is subject to the customer’s limits.
We think this is one of the most commercially interesting developments. For supported products, it creates room to offer an entry-level experience powered by AI the customer already pays for. A business could then charge for specialist functionality, proprietary information, team collaboration or a wider service.
That could make focused products easier to try and support different pricing models. It does not make AI usage unlimited, so the product still needs to explain what happens when a customer reaches their allowance.
4. Decisions API: intelligence built into everyday workflows
The Decisions API accepts text or images and selects from predefined answers. OpenAI announced it in limited preview for tasks such as classification and routing.
This could be useful in products that handle large volumes of incoming information. Requests could reach the appropriate team sooner, documents could enter the right review process, and routine cases could be separated from those needing attention.
The value could come from handling more work without a corresponding increase in administration, or making a product faster and easier to use. Start with a decision where the possible answers are clear and you can check whether the system chose correctly.
5. MCP Events: services that respond when something happens
MCP Events lets users subscribe ChatGPT to changes in connected systems and specify the response. A new task or document update can trigger follow-up work.
This opens up possibilities for services built around ongoing monitoring. A product could signal that new information has arrived, prompting analysis, preparation or a request for human review.
For customers, the benefit is less checking, chasing and remembering to start the next step. For a product business, that could become a valuable ongoing service.
The workflow needs to be useful when nothing unusual happens, too. Customers should be able to choose what deserves their attention, understand what has run and stop monitoring easily. Repeated events should not create duplicate work or a stream of unnecessary notifications.
What’s worth building now?
Taken together, these announcements offer more ways to build useful software and put it in front of customers. We see room for businesses that understand a particular problem and can build a dependable product around it.
A sensible first step is to choose one customer task, establish what a better result would look like and test the relevant capability against it. Check access conditions and usage costs before making them central to the business case. Then decide what belongs in the first release and what needs further evidence.
Good user experience, reliable integrations and clear controls still matter. Those are the things that turn a promising capability into software people choose, trust and pay for.
DabApps is an OpenAI Select Partner, with more than 15 years of experience designing, building and supporting web applications and mobile products. Through our AI consultancy, we help businesses assess opportunities, take new products to launch and add useful AI capabilities to software they already have.