AI Copilot in Power Apps

In the previous blog by Senior Consultant Steven Brusse, not even two months ago, he wrote about the influence of Artificial Intelligence (AI) on low-code software development. In it, he raised the question of whether it will ever be possible to have applications developed by an AI model. A few weeks later, Microsoft already gave an answer to this question with the introduction of the AI Copilot in Power Apps. With the AI Copilot, Microsoft shows what is possible when a generative AI model like ChatGPT is integrated into a low-code development platform. Anyone can now get started building an application, simply by describing the desired specifications in human language through a conversation with a chatbot. With modern low-code tools, programming experience was already no longer required to build working applications, but without specific tool knowledge you still didn't get very far in practice. Could that now be about to change, or is it still too early to draw that conclusion? In this blog you'll read what capabilities the new AI Copilot currently offers, what its limitations are, and how you can request a trial version to test it yourself.

Signing up for the trial version

Before we get into the capabilities of the new AI assistant, it's important to mention that the AI Copilot is not yet available to everyone at this time. The new AI Copilot features are not yet intended for production use and are currently only offered as a limited preview. You need to submit a request to Microsoft to be able to use a trial version. A condition for participation is that your environment (Power Platform Environment) is hosted in the US and that AI Builder is enabled on the environment. The request for the trial version can be submitted via an online form.

What can AI Copilot do?

The new AI Copilot is a smart assistant that you can put to work for you in various ways when developing apps (with Power Apps), flows (with Power Automate), and chatbots (with Power Virtual Agents). In this blog I'll limit myself to the applications of AI Copilot in Power Apps.

• Building apps through a conversation
• Adding the Copilot control to apps

Building apps through conversation

As soon as Copilot is activated in your environment, on the start screen in Power Apps you're asked to describe what kind of app you want to create. Here you can describe, in human language, what kind of information you want to collect, register, process, or display. Based on your description, Copilot generates a new table in Dataverse with the back-end logic to automate the requested activities. You're then shown the structure of the new table along with some sample data.

You can modify the new table, adjust columns, or add extra columns by describing the desired changes. To help you along, suggestions for possible follow-up questions you can ask Copilot are provided after every new question. You can not only ask Copilot to change the structure of the table, but also to generate additional sample data.

Once the table meets your requirements, you can finally ask Copilot to create an app. The app that gets created is a responsive canvas app with a start page and an input page. On the start page, the new Copilot control has been added (see the next paragraph), which lets you directly ask questions about the data in the app. The second page contains a list displaying the existing rows from the table. Here you can edit rows and add new ones.

Adding the Copilot control to apps

The new Copilot control is an AI assistant that developers can add to apps for end users. With this control, end users can gain insights into the data stored in the app through conversations in human language.

The Copilot control is automatically added to your app when you create the app with Copilot. In addition, as a developer you can add the Copilot control to existing apps that already use tables in Dataverse. Here, in the control's settings, you can choose which data questions may be asked about.

With this new control, end users can ask highly targeted questions about the data in an app, making the app itself more user-friendly. Apps can also be made simpler, since other controls for, for example, looking up, displaying, and filtering data are no longer needed.

What are the current limitations?

Although the new AI Copilot features are currently only released via a trial version, they already give a nice picture of the future of low-code software development with Power Apps. Based on the current capabilities and limitations, however, the new features are not yet usable for production applications. I've listed a number of limitations below:

• Copilot can only create a canvas app, not a model-driven app.
• The apps you can build with Copilot are very simple, because they're linked to only one table in Dataverse.
• If you later want to modify a canvas app and extend it with extra functionality, you can't use Copilot, so you'll still need someone with specific tool knowledge after all.
• The idea behind the new Copilot control is very powerful, but you can only select a single table for the control. The questions you as a user can ask via the control are therefore limited, since data is normally stored across multiple tables.

Conclusion

The AI Copilot preview proves that it's possible to develop apps with the help of a smart chatbot without any prior knowledge of Power Apps. Due to the limitations that currently apply, Copilot isn't really usable in practice yet. That will undoubtedly change in the near future. Microsoft has announced that it's fully committing to AI applications within its existing cloud services. For that reason, I expect the capabilities of AI Copilot to be expanded quickly. On the other hand, I don't think experienced app developers with programming experience and/or specialist tool knowledge will become redundant because of an AI-driven chatbot. For complex applications that integrate with external systems, these kinds of specialists are still needed, and that's likely to remain the case.

Want to know more about low-code?

Want to know more about the possibilities of low-code tools, or do you have a specific challenge that a low-code app could help solve? We're happy to help with advice and to think along about the possible applications of low-code for your specific situation. Get in touch with one of our account managers to schedule an introductory conversation.

Get in touchSchedule an appointment

Written by Steven Samuels Brusse,
Senior Consultant