"No Code" AI Apps Generators

Testing Some Conversational User Interfaces in Application Generators.

SSTC has been engaged to provide guidance for a no code platform development. Our expertise is really in the Low Code space, so it seemed appropriate that we scanned some of the leading No Code development capabilities. These are some of our findings.

Bolt

Bolt was amazing. I am doing a coast to coast cycle tour in a cycle of weeks, and had needed to find somewhere to park the car. With a very limited prompt, Bolt made a very good guess at an application to provide that service to visitors to Nottingham.

Bolt allows you to interact with a chatbot to discuss the application. It builds the application in React right in front of you. You can verify the application in the same window. You can deploy the application from within Bolt or you can take that code somewhere else.

I also often play cards with my dad. I asked it to create a card deck and table and it brought in the appropriate libraries to show cards on a green baize background.

Pro’s

It made a really good guess at layout, buttons. It looks like a modern website.

It is a super compact IDE that includes the conversational interface.

Con's

I get less of a sense of the flow through an application.

There seems to be no progressive immersion into the complexity. You are either at the chatbot or in React.

Zapier

Zapier bills itself more as a workflow automation tool rather than an application generator.

It offers a similar Chatbot interface to describe and refine the application.

It shows you a canvas, displaying the design assets: Pages, Tables, Notifications.

You can test out the pages and the deployed application right from the Integrated Development Environment.

Pro’s

The development environment feels a little more gradual in its exposure of complexity. It feels a little closure to the Microsoft Power Apps experience.

Con’s

Much like the Power Apps Experience you feel like you are in the baby pool, knowing that you will have to brave the chilly air of skills development to get into the big pool.

RelevanceAI

Relevance bills itself as an environment for developing Agents.

This is not the first time that the word Agents has been used to describe a set of programs. I remember, back in the ‘90’s when people were trying to represent resources on a shop floor as agents getting rewarded for removing “Urgency” in work packets as a way to optimize the flow of work. At the time, I thought using the word “Agent” was a way to push all the questions to the other side of a wall of assumptions. By calling a program an agent, we were tempted to assume it was somehow radically different from the shop floor optimization we were familiar with.

Relevance seems to be building scaffolding to embed a Large Language Model and connect it to tools and abilities with which it can take actions in the world.

I am still propagating my blog posts to social media channels by hand, so this seemed a good use case to explore.

I used its content writer to prepare a post on exactly the same topic I am writing on in this article.

It is very well written, but we seem to have different opinions on the tools in question.

I also tried out the Social Media Propagator.

It did a pretty good job of summarizing and tuning the content for each of the platforms.

I did this by taking a copy of one of the templates. That copy has placeholders for Tools like Get Linked In Post and Post to Twitter.

You can edit or define these tools within Relevance.

What becomes clear quickly, is that you are no longer in the baby pool, and you are no longer in the shallow end of the big pool.

Pro’s

I made good use of agents from the existing library of templates. I like the Core Instructions and the Flow part of the agent definition. This seems like my first conversations with a client as an architect trying to elicit the desired behavior. There appears to be good onboard learning content.

Con’s

It seems that you would need to invest a fair amount to be competent in the tools definition piece. In my experience, the people that are really good at requirements solicitation tend to be a different set of people than those that are really good at picking up a new tool.

Conclusion

The conversational UI to Application is pretty amazing in both Bolt and Zapier. Zapier probably lets a less technically competent user feel in control of the application being created. Bolt gives a more technically competent user a springboard for development. The agent creation in Relevance comes with some useful examples and great learning resources, but the tool takes a technical mindset to get full use of.

Google Sites

Report abuse