Harnessing Low-Code Development and Generative AI

Indu P. Chaube at UXDX Community: AI, Low-Code & Designing Quality. Video: https://youtu.be/kMkmiO53Jl0

Readable transcript: edited from the recording's captions for readability (fillers and false starts removed, punctuation and section headings added). Wording is the speaker's own. Timestamps are positions in the video. Names marked [?] could not be verified against the audio.

Low code plus generative AI

[00:00:07] Thank you. Hello everyone. Today we are going to talk about UI/UX design with low code and generative AI. What does that mean? Let's explore. This is a fun journey today you are going to go on with me. There is a low code platform, a GenAI agent, and how this particular GenAI agent is going to help in low code platform development. At the result we are going to see an AI-driven low code platform.

[00:00:34] So as a brief, earlier a low code platform simplifies the development with minimal coding, drag and drops and configuration tools. On the other hand GenAI is going to assist in creating wireframes, assets, query data processing in real time. So if they both get connected in real time they produce an AI-driven low code platform where the user can personalize and set their preferences in real time. That is the goal of this discussion today.

[00:01:06] So into the traditional software development, how it has been done: it is totally static, and it depends on what the user inputs and what is the output. There are no preferences of personalization in traditional software development. But in the modern, like right now, we know a low code platform that enables the user to create and personalize their interface in real time. Not only personalize, they improve the user experience also.

[00:01:38] On the other hand, we are seeing generative AI and agents right now in the industry. How are they impacting the whole of software development? They are producing content in real time. What if this content gets delivered directly to the low code platform, where this agent could be adapted as a part of the low code platform, and then the user starts personalizing their experience according to their need? I mean, all the data processing is going to happen in real time. That is what the future is going to look like.

Training the agent on a specific data set

[00:02:17] So low code platforms streamline development by reducing coding complexity and enhancing user experience, and on the other hand generative AI is going to add real-time interactivity by dynamic adaptation of design and features. AI trained over a specific data context: let's suppose we are developing a particular user interface on a specific data set. What we have to do is we have to train that generative agent on that particular specific data set, so that if the data set is growing, probably the structure is not going to change too much. So by leveraging the capability of GenAI, producing a query in real time, and that connects to the UI in real time, it reduces a lot of time and gives an improved user experience to the user. So this collaboration is set to revolutionize how we think today, bringing transformative change to software development, digital design and user experience. That is what I am expecting in future.

Challenges in UI/UX design

[00:03:28] And challenges in UI/UX design. If you see traditional software development, it is too time consuming, a lengthy procedure, and it slows down project completion. This can frustrate developers and stakeholders. We know if we go through the whole software development life cycle from end to end, probably any piece of information getting into reality is going to take a lot of time.

[00:04:02] And then we know this software technology is also creating a bottleneck. A lot of technology is going to happen in the coming future, so skill barriers are there. GenAI is going to help us in improving those barriers. That is, software development languages adoption is made easier, right, because it compiles the code and it generates the code for us, so that we can build our feature set into the products very easily.

[00:04:39] And one more challenge is that connecting from business units to the developers, it is also kind of lacking a lot of collaboration. When I'm saying that the low code platform connects with the AI and it is producing results in real time, probably this collaboration can be more simplified in future.

[00:05:00] And we see difficulties in creating prototyping also. So if AI and low code get connected, probably we can prototype a particular thought process into reality in very little time. So that is also getting solved through this solution. And lack of adaptability, because we cannot personalize or set user preferences in real time. The user doesn't know what they need, and then it goes into a deadlock life cycle as to what exactly needs to be delivered. So these are the challenges. These challenges are going to be solved with connecting the dots between low code platform and AI.

Use cases across industries

[00:05:51] Use cases, likewise industry. We see huge cases kind of increasing day to day. There are a lot of industries, likewise e-commerce, healthcare, financial services, marketing and education. Their requirements are getting changed too fast and software needs are getting changed too fast, right? So just for example, let's suppose e-commerce. There are many, many wide ranges of products coming into the market and their way of selling is also getting changed.

[00:06:30] Similarly in healthcare automation, if you see, there is a wide range of opportunity for developing new features, but if you follow traditional software development, probably we are going to delay the delivery to the end user. In that case, how AI is going to help us is kind of building a low code platform, connecting dots with the AI tools. If I say AI tools, it means any AI agent that is totally trained on the specific data set is going to produce better results in an efficient time.

[00:07:10] Similarly for the financial services, marketing campaigns and education and training, it is going to produce the content according to the user need, and it is going to represent the content the way the user wants. Let's suppose particular information is going to be seen in a table, or is going to be seen in a chart. We know generative AI is capable enough to produce code for charts and tables. And what we have to do is we have to connect the dots, meaning data to the GenAI, and then GenAI produces the visualization that directly interacts with the user interface.

[00:07:50] You can imagine how powerful this is. If we connect the data and data compilation, at the same time we create a user experience in real time and deliver it to the platform so that we can serve our customers and users in real time. How useful it is.

The advantages

[00:08:12] And there is a lot of advantage of it, like faster development, because we are not doing a lot of development behind the scenes. Everything is getting done in real time through the platform. The platform that I'm saying is an AI-driven low code platform, delivering the customer or partner needs, or user needs.

[00:08:32] It is cost effective also, because everything is happening in real time. Though training an AI agent is kind of expensive, but it is a kind of very, very much standard process going to be adopted from the industry. So it is cost effective.

[00:08:55] And accessibility for non-technical users also. Leveraging low code and AI together, we do not need to train someone in the skills, right? Even a user or someone who doesn't have technical knowledge can build a feature on the fly from the platform itself.

[00:09:18] Improving collaboration: in this case developer, designer and business owner can sit together and can craft the software need according to their requirements. Building features, security, monitoring and et cetera, we have to build one time, and one time for many requirements, many software features into the platform. So all these together, if we have that kind of capability, probably we have to focus on innovation, and then we can provide a wide range of feature sets to the customers and partners in very, very efficient time.

The motherboard analogy and conclusion

[00:10:05] So apart from this discussion, I'm going on to the conclusion of my talk. In the conclusion, if you see, when I was saying low code, probably everyone knows now what a low code platform is. A low code platform is kind of pre-built software infrastructure, like a computer motherboard, right? So a computer motherboard is kind of a pre-built low code platform where we can connect a monitor, a wide range of monitors, mouse and cameras, right? Maybe we can connect many monitors, not only one monitor, according to our need. Similarly, low code platforms enable us, whatever technology is going to happen, we don't worry too much. We have a low code platform where we can plug a generative plug-in or any other plugins in real time.

[00:11:08] The second part of it is kind of, we are promoting innovation by doing this, because users and partners or anybody who is using the software platform can configure software in real time. So that is kind of liberty to the user, who can plug and play with the platform and craft their software according to their need. What we have to care about is we have to get data, and data can be delivered by connecting the dots between AI and low code.

[00:11:43] And the collaboration prospect, if we see, it is going to simplify designer, developer and AI together, right? So once these things happen together it promotes a lot of collaboration in real time. That is also saving a lot of time for the business to go to market.

[00:12:05] And emerging technology. We know today GenAI, tomorrow something else, right? How can we connect new technology to the platform? Only when we have a low code platform where we can plug in anything in real time. That is the power of seeing a low code platform along with AI, that it can have a future and connect different kinds of technology in real time, when the time comes.

The best way to predict the future is to create it

[00:12:38] So at the end, there is a very wonderful quote. Everybody is thinking about AI and how we are going to use it. In my philosophy, AI is the most powerful tool nowadays in the market, and adoption is very fast in the industry. It can be adopted even faster if we have something that is called low code software infrastructure, where we can welcome any kind of technology, not only the AI or AI agent or something else that comes in the future. But for now, building a low code platform and connecting it with the technology that is available in the market, that is kind of key.

[00:13:23] So the best way to predict the future is to create it. Now the time is kind of creating a platform that supports anything in the future. So working now for the future is kind of low code, that I'm promoting today.

[00:13:41] And low code and GenAI together is fun. Likewise, if you see, we want to play with the software, we want to see the information the way we want, right? So it is fun, we can use software development as fun, right? And then later on whatever we are creating becomes a future. So low code and GenAI together is fun and it is a future. That is what I'm promoting in today's talk. It is very minimal, because it was a short talk. I thought I'll be creating some prototype on top of it, but I just want to cover how GenAI is impacting the world and how low code platforms come together along with AI and impact the future in an even easier and faster way. So with this conclusion, thank you so much for joining me. Any questions that I can answer now?

Q&A

[00:14:50] Host: Excellent. Thank you very much, Indu. It's a really topical conversation, so I'm hopeful there'll be some questions out there from people, but I'd like to get started, because one of the topics that you brought up was that people want to be able to change things themselves. As you said, you gave the example of data visualization. I found platforms tend to be opinionated. So how do you marry those two if you're stuck with the opinions of the platform and you want to do something different?

[00:15:22] Indu: Yeah, that is a very good question. When I say low code, low code is a liberty for the industry to welcome anything. That was my goal while I was talking about low code platforms. It means that we can welcome any plug-in into the infrastructure. This is not the first time it has happened. There are many companies that are developing software like that, that can welcome any kind of additional plug-in to the infrastructure and let them do what they want to do. At the same time they could enable an interface so that the platform can connect with the new plugins. GenAI, I'm saying GenAI is even more powerful. It is not a plug-in in that case, it can do certain things in real time.

[00:16:14] Indu: So visualization is one of the examples. I can say, only two, three, four, five fields I want in the UI, right, and I want to visualize that UI in many ways, like counters, like a table, like a line chart, something like that. How could I do that? GenAI can create a real-time query on top of the data and can produce it to the interface. Are you connecting with me, whatever I'm trying to say GenAI is?

[00:16:46] Host: So to summarize, or to kind of get my understanding, it's from a very low level. So your platform would really be like a database level and then you can do whatever you want on top of it, versus where my opinion was your platform could be a financial system or it could be a CRM, which are much higher up on the stack. So your recommendation is, I guess, lower platforms that then enable people to do what they want with AI.

[00:17:15] Indu: That's right, that's right. And also one thing that I want to repeat again, the agent-based GenAI. Agent-based GenAI means it is trained on a specific data structure. Let's suppose you are a financial company and you are producing a lot of financial data, and for financial data you have a lot of reference. What do we have to do? We have to train GenAI on those data, those structures, so that when a query happens in real time those data get summarized accordingly in the easiest and fastest way.

[00:18:00] Host: And do you think that that's been the approach for the last few years, of okay, if you've got a niche, train it on your niche and the data will improve. That's with the general models getting more and more intelligent with each iteration. Do you think that that will remain, or do you think that the general models will become so knowledgeable?

[00:18:16] Indu: Oh, that is their industry, it is moving, it should be more and more smart day to day. That's why I said a smaller data set and a well trained model can do even much better. Whatever the user needs, they can produce the result accordingly. It is kind of writing a SQL query into the service layer, or telling GenAI write a query in real time and send the data back exactly as it is needed. You know, right now people are talking about GenAI or agents being biased and biased, but it depends on what kind of data set that particular GenAI agent gets trained on.

[00:19:05] Host: And how do you deal with the challenge that a lot of people face, and I've personally experienced it, that it can be great and amazing and then the next day or the next query can be completely wrong, because it just, I don't know why it does it, misinterprets something?

[00:19:24] Indu: Yeah, I got your point. If that is the case, that's why I said do not use GenAI as a part of the solution, use GenAI as a plug-in for the solution. So whatever result it is producing, review them. If that result is okay for the data, right. Let's suppose I ask GenAI, I write a query to produce this chart, right. GenAI goes and writes a query onto the structure and returns data into the chart. But also it is going to flag you out at each and every step, where the query could be written and then data could plug in with the charting library, right, where the user can see. And if the user can feel that the query is not right, that could be corrected. A low code platform enables you to correct that query in real time and then save it and produce the result accordingly. Not only save for this particular time, save for the future also, with the naming alias. And this plug-in is ready to call again and again. And if you want to produce new plugins for different kinds of data visualization, ask GenAI to do it again. If your review is kind of fine, just plug that particular feature set with the interface in real time and you are done.

[00:20:49] Host: So in your model, and correct me if I'm wrong here because I'm trying to interpret as you're saying it, it's still a developer prompting the AI on how to create these plugins, versus let's say a marketing person or a person in operations. It's still a developer who has enough knowledge to be able to validate and determine the quality of the code. Or do you see longer term it becomes that marketer, it becomes that operations person, that finance person?

[00:21:18] Indu: The future, what is going to happen, I don't know now, but one thing I told earlier is developer, designer and business owner can sit together and collaborate in real time. So for a specific need, someone tells the business need and someone is sitting there prompting the agent that hey, this particular result I want. If there is anything like data related problems or query related problems, the developer can figure that out. At the same time the designer can review in real time, is this particular thing what users want or something else? If not, the same life cycle could happen again. The prompting keeps going till the produced result is whatever the requirement says. You got my point? That's why I said it enables a lot of collaboration in real time, other than discussing in the meeting, going back, doing code and coming back for the review. That is the process getting eliminated by enabling low code and GenAI together.

[00:22:31] Host: Brilliant, yeah, I love that collaboration approach. I'm going to ask you, because you just said no, you can't guess into the future, but where do you think, because from our conversation so far I'm thinking what's the value of the platform going forward if the AI is that intelligent? Do you even need the platform, or can you just have AI generating these custom apps for you as needed rather than plugins?

[00:22:56] Indu: Oh, that is a very, very good question. See, when I'm saying platform, likewise if I say build a house, I want my house to be in a different structure, but what I want to change is, I want to change the color, I want to change the light inside the house, where they get fitted. I want to change the color of the table. I want to change the monitors and printers and all. GenAI is going to facilitate those things, not how my foundation is getting built. So GenAI now or in future, I believe it is just a tool that can facilitate something that could be happening in the platform on demand. You are getting my point?

[00:23:52] Host: I mean, I do not want structure within constraints, I guess would be, because if you ask AI to do something and no constraints, you could end up in a completely different direction. Would that be a correct summarization?

[00:24:06] Indu: That's right, that's right. See, I want to do certain works from AI the way I want. I want full control by leveraging AI.

[00:24:22] Host: Yep. As somebody I heard say, instead of vibe coding they call it mom coding, because as a mom you have to keep an eye on your kids. You give them a suggestion and then you have to stop them as soon as they start going in the wrong direction.

[00:24:36] Indu: That's right, that's right. An interesting example.

[00:24:39] Host: Because I know personally sometimes you prompt it and it goes in a completely different direction to what was intended.

[00:24:46] Indu: Yeah, yeah. See, where there is a lot of liberty we should have control. We can party the whole night, or we have someone to control, hey, till midnight all done, your party has been done, and someone needs to take control. That is why I promote a platform. At least, I'm not saying no code platform, I'm saying low code platform, where we have full control. If something needs to be tweaked, that can be done right away.

[00:25:22] Indu: Okay, and then let AI help. AI is a tool that can help us in real time, but not take control of what we want to build on top of it.

[00:25:35] Host: So using AI as the tool rather than just letting it run, letting it do anything.

[00:25:40] Indu: Yeah, I want control. I want full control, because whatever I'm using AI for is for my need, not according to AI's need, right? AI can suggest anything the way they want, or it suggests only the things it is getting trained on, right? If data is bad it could be more biased. So I want to train on my data so that it could not be biased. Even if it could be biased, I want to have full control. I want to watch it, and then if it is producing the result according to the need then it is going to be enabled into the platform, only when it is all verified, what are the results that get produced, when human intervention is needed for the feature set to go in.

[00:26:38] Host: Brilliant. That actually brings us to time. So thank you very much Indu for sharing about, I guess, where the future of AI and low code, not no code platforms but low code platforms, fit in.

[00:26:50] Indu: That's right, yeah. Thank you very much. Thank you everyone. Thanks.