New Crowdbotics Capabilities Are Gen AI Game Changers

Crowdbotics uses AI to accelerate every step of the software development life cycle (SDLC)–starting with ensuring every project is built on the most solid foundation possible. Perhaps the most pivotal step in the SDLC–the creation of high-quality product requirement documents (PRDs)–has long been an underinvested step. Given this, it’s no wonder that the number one reason software development projects fail is poor or missing project requirements.

Crowdbotics began changing this dynamic with the introduction of PRD-AI at the beginning of 2024. We use an LLM to ensure that every PRD comprehensively expresses the business requirements. Furthermore, this approach ensures that the requirements are detailed enough to ensure developers clearly understand the stakeholder’s intent. It turns out that bridging the historical chasm between business and IT is a perfect application for generative AI.

Also introduced in January, our platform uses the PRD to identify reusable code modules. Customers can curate a private library of code, and they have access to public modules curated by Crowdbotics. Once a module has been identified as being useful, the platform automatically installs the module into the project repository. By systematically utilizing reusable code, the Crowdbotics platform is able to deliver enormous productivity boosts for teams building large portfolios of apps, and this is all made possible by using AI to match requirements in the PRD with the functionality associated with cataloged code modules.

It turns out that the context contained in a PRD that enables matching code modules can also be used to power AI-based recommendations on architecture, and even more impactfully, it can automatically generate new code. This is exactly what we’ve done with two new platform capabilities: Technical Recommendations and Application Code Generation.

Technical Recommendations

Once a project’s requirements are locked, the platform utilizes AI to breakdown each feature requirement into layered functional specs across Backend, Frontend, Data Schema and 3rd part integrations. The new Technical Recommendation feature — using AI — then provides holistic recommendations across these layers. For instance, for each and every requirement in a PRD, Technical Recommendations will provide a roadmap for the development team on how to implement that feature. It provides a recommendation on how to segment functionality between front-end code and back-end services. It will also suggest the proper data structures to support the feature.

Technical Recommendations – taking into account all of an organization’s context – will also generate a list of suggested services and third-party APIs that may be required to implement the feature. For example, if the requirement in the PRD is for a login dialog to support user authentication, Technical Recommendations will recommend a cloud authentication service to support this requirement. If the enterprise’s standards mandate that they build their apps on Microsoft Azure, then Technical Recommendations will suggest Microsoft Entra ID as the proper service to support authentication. If, on the other hand, the organization is standardized on Okta, then that service will be suggested by Technical Recommendations as a reminder to the developer of this corporate standard.

Often, Technical Recommendations will suggest several third-party services if they all fulfil the requirement. For example, the app has a requirement that includes sending notifications to users, Technical Recommendations will suggest Azure Notification Hubs as well as Firebase and Twilio as possible services to use to implement this feature. Again, if the customer has expressed a preference for Azure, then only Azure Notification Hubs will show up in the Technical Recommendations.

Another interesting aspect of Technical Recommendations is that it can highlight features or functions that will be reused in the implementation of other features in the PRD. This ensures that features only get built once, no matter how many times they are reused and no matter how many developers are collaborating on this project.

At its core, Technical Recommendations uses AI to accelerate the planning phase of a development project as well as to remove risk by enforcing standards and consistency.

Example of Technical Recommendations embedded in a PRD.

Application Code Generation

With all the rich technical context generated through Technical Recommendations, the platform now has enough context to be able to generate as much as 50% of the code for an application. And even cooler, this code generation can happen in bulk, before developers even start working! This is precisely what the platform’s Application Code Generation feature delivers.

Application Code Generation uses all of the contextual information from your PRD, from your Enterprise Context (see today’s other blog here), and from the associated Technical Recommendations to generate much of the code required for your app. It takes care of much of the toil and commodity code such as generating code for CRUD data operations (i.e. create, read, update, delete). It can generate all of the code for calling third-party APIs. For instance, if Technical Recommendations suggest using Microsoft Entra ID for authentication, Application Code Generation will create all the code required to call this service and to wire it into the login dialog. It can generate code for common functions that were suggested by Technical Recommendations. It can generate code to expose service endpoints. There are myriad areas where Application Code Generation will fill in the blanks for you.

It’s important to note, though, that the blanks being filled in by Application Code Generation are not just associated with commodity code–often referred to as toil. It can also implement crucial business logic such as data validation rules and work flows as defined by the PRD. Because the Application Code Generation feature takes advantage of the broadest possible wealth of context, accuracy and code coverage are maximized. This approach dramatically lightens the load for developers. Essentially, the combination of features starting with PRD-AI and including Technical Recommendations and Application Code Generation enable you to go from a textual description of an app to a 50% filled code repository in the time that most organizations take to just schedule a planning meeting.

Example of code generated by the Application Code Generation feature.

SUMMARY

As previously noted, with the current generation of LLMs, we are able to generate as much as 50% of the code for most business applications. We are confident that, in the near future, we will be able to generate the majority–if not all–of the code for many apps. Our top-down approach to code generation, which we refer to as CodeOps, is the key to this vision. Our approach always starts with high-fidelity, AI-driven requirements. This sets the foundation for us to be able to eventually generate feature-complete applications. The approach is proven, and our platform is well positioned to help you capitalize today and well into the future as the underlying AI models continue to advance.

We’re excited about delivering these latest advancements to the Crowdbotics platform. We believe that these features, when taken together, are game changing when it comes to software development. Most of all, we’re excited about the acceleration in productivity our customers get with these new features. We’re building a platform designed to help you and your team build apps faster and with less risk, and these new capabilities are a huge leap forward in helping you succeed.

Go from Idea to Code with AI

Schedule a Demo