 Hi everyone. Hope you had a nice coffee break. I am Rupesh. I work at Quilbott and let me start with talk about pigeons. So, whenever I have a creative block, I am looking outside my window and I seldomly find a pigeon looking back at me and saying, bro, do you need any help? Whenever I went to talk to this pigeon, he just flies away, right? Such a strange behavior from a round little bird with a weird color palette. But some guy a few thousand years back looked at a pigeon and he saw gold in it. He saw global business. He saw opportunity. So he got a bunch of pigeons together. He trained them. He got them lifting weights, gave them some protein shakes and he built up pigeon post. So this is like the first messaging platform that really sped up long-distance messaging. During those days, people were all about face-to-face communication. They really like to communicate face-to-face and long-distance communication just had picked up. So if you are sending a message to say from Bangalore to Mumbai, it is to take a thousand kilometers and 16 days of horse-riding effort. Somebody will go to Mumbai, then pick up that message, come back and it should take 32 days for communication. But pigeons, as tech, reduce that gap by almost 10 times. So it will take like couple of days for the pigeon and they use innovative relay mechanisms to send messages, right? So while we are talking about pigeons, we need to understand what really happened here. There was a need to send messages on long distances and there was this nice, beautiful bird as tech which evolved into a product, right? Now if I look at this example, a very simplistic example of need plus tech into product and I scale it, do we use pigeons now to send messages? I hope not. Do we use quill and paper to write on pages, right? We really don't use this tech anymore. So if I upscale this equation and say there is this persistent need for all of us, right? We still communicate, yeah? We still send messages. So this need, which was there thousands of years before, is going to continue to exist thousands of years later, right? I hope so. And this persistent need, if we add evolving tech. So tech, pigeons, as a tech, was really low tech. It has like short firm. You can only do 30 characters at a time. But if you evolve it and meet it to the current standards, you really get something by periodic disruptions, you get like an evergreen product. So an evergreen product is something which lasts for like centuries, right? So food, for example, is an evergreen product. Everybody eats food. They have been eating food for thousands of years and they are going to continue to eat so. Clothing is like an evergreen product. People used to wear clothes and I hope so they'll wear in the future as well. So if I add AI to this tech, which is AI basically disrupts all of your existing workflows. So the persistent need, if I add AI tech to it, we still get an evergreen product. Now, hence when I was working at Peelbot, I was working for around 18 months, one and a half years, and I was working on a lot of AI projects. But I felt that the traditional methods were not working out. So I used design thinking. I was using a lot of brainstorming sessions. And hence I thought I didn't have something different, something which is a little bit evolved to handle this. So I designed this framework. It's just fresh, like three months, and we are testing it out. But by scale, I want to transcend tech by AI-driven design. I want to use the traditional practices to leverage AI as much as possible. And I just want to glimglue a game today. So if I have to deep dive, it'll take hours. But I'll just glimpse of the framework. So what the first thing you need to do is see the evolution. It's a little aspirational, but you want to see what has happened over the years. And when you're introducing AI into it, you will see the change, what benefits we're adding. Then you need to capture the constraints. When you are working in an AI environment, there are so many constraints that we need to look for. And you need to capture them all. Third is adaptive AI design, where we are just leveraging AI to design as much standard and non-interface as possible. And then you need to iterate and validate and learn and evolve. So I kept some of the design thinking principles the same, but the initial three is something which we can focus on. So seeing the evolution. Now, at QuillBot, we are also an LLM, but we are focused more on paraphrasing and, say, we have a suite of tools of grammar checker summarizer, co-writer. But when we add Gen AI to it, to our features, to our products, the workflow is really getting disturbed. Somebody is doing a common task of sending an email, and then we are generating AI to it. It really is not in their common things to do. So how do you address this? So we need to ask, is AI simplifying your task? Is it reducing the complexity out of it? Is it reducing the time taken? Is it reducing the number of clicks? Or is AI adding value? Is the output significantly better than you would have done manually? Are you going to some other platform to do this? Does it add value? So if you want to see the evolution, you need to identify three things. You need to find that persistent need which was there, which is going to exist, which is going to be continuously evolving. You need to see the tech evolution. You need to know what the first-gen flow was. You need to know the current-gen flow. And then you need to see how you can transcend this tech, evolve this tech by using AI. The last thing is, you need to be clear-cut on what this value add from AI is. Then you can just spread up and get prepared for your AI project. So this has helped us in identifying the value add and the preposition. And it has also helped us users to streamline their workflows. So one such tool is our paraphraser, which basically rephrases your text. And the beauty of this is, you can use it anywhere. So no need to go to chat GPT. You can just use it in Gmail, Twitter, LinkedIn. It's just an extension. And it gives you basically spoon feeding. So you don't need to think much. And you can just add some different modes of fluency, simple, creative. It works anywhere. And it has different modes. So it will basically help you ease into your tasks. So the next thing is capturing the constraints. Now PigeonPost, which were there centuries back, has now evolved into PMail. So PMail is a communication platform. And it uses email. And it is just like our pigeons found some new jobs. But my CXO comes to me and says, boss, PMail is now outdated. We need to evolve this tech. And we need to call it P, the everything app. So I went back to my CEO and was like, what is this everything app? And this basically means that the app has to do everything. And it was so difficult for me to bridge that from email to I'm going to do everything else. And as a designer, I straight away jumped into Figma. That should not be the case. I need to understand what the business means by everything. The first thing you need to do is have a clear understanding of what the business wants from this project, what the business is looking for. How can I address his needs? I was talking to my product manager and he was giving me this example, that whenever there is a transfer of information from the top level to the most junior designer, you lose 10%. So whenever you are having a vision, the CEO is having a vision, ultimately what you develop is just 10% of what it is. So how do we capture this? So involve your stakeholders, your CXOs, your clients, your managers in your brainstorming sessions. Like grasp as much as possible. And I've found guilty of that. Like in many times I like my CEO to sketch. I make him do crazy eights. I'm making him do how my tweets. And I know I'm taking his valuable time, but I'm clear cut on what information I want. So this is a very major constraint which we really miss out. We just have a PRD which is live and we just work on requirements, but rather than caption the emotions in which the vision was there for the product. The next thing is basically tech. So AI is really tricky when we want to understand this tech. When we are looking at AI tech, there are so many constraints of what is the cost factor? What is the performance factor? What is the latency? How do I factor this into my workflows? How do I make it easy for the users as well as the developers as well as the designers to understand how is tech being used? So many a times I do my entire flows. I have my everything done, prototype done. It's research is done. And then tech will say, boss, the latency is so much that you can't really use this. I need to add some additional loading screens. I need to reduce the prompt count from say 50 to five. Like how do I address this? And it goes, I go into a cycle of really redoing some of my screens. So we can avoid that by again involving tech into our discussions, get them into brainstorms, do feasibility as much as possible. Next thing is time. So how many of you have been asked on Monday to deliver something on Friday? Like there's a ticking time bomb always on our heads. How do we address this? So don't really try to reinvent the wheel when you just have a week's time. And also don't jump into Figma that same day. Use the time to do a lot of due diligence, lot of research, and just scale it, right? If you have two weeks versus you have two months, you need to know which deliverals, which topics to address, right? That's why we're time. And the most important thing is context, right? Make sure your AI is generating those contextual responses. Really make sure that it is adding value. And together with business, tech, time and AI, you define your MVP. This is the most important thing for AI project, is defining that basic MVP. It may be as simple as this chair with a little bit of AI, but it is useful for setting. But don't go for that gaming chair in the first attempt. Just try to focus on your basic MVP, your basic change, your basic driver that's gonna hook your users on, right? What about branding? Should we do branding? I think you should just leverage your ecosystem of design systems, your existing colors. Just add some sparkles to that icon to make sure that it is a AI product and just be done with it. I don't think you should focus too much on the branding on the MVP. Test your MVP, the users. If it is found successful, then you go for a rebranding exercise because working on AI project is really costly, even for all streams of channels, if designers, developers, it really takes an effort. So when you're capturing the constraints, try to define a proper MVP. See it's at the context and just try to build a simple AI solution. Don't go for complicated things in the first attempt. Just go for something very simple, something incremental that will be very useful. Here is one feature which it's like ifs and buts, they're in there, where we try to use a feature plus quick compose to write your emails, taking a context which is a subject or some body copy. So it's like a one button compose and I really try to reinvent the wheel. That's why we are sort of differing a bit from our goal, where instead of just giving them a chat interface, which is pretty standard, like everybody in Gen AI is using a chat interface. Now when I look at a chat interface, I find it difficult to extract value out of it. I had to go through the text. I can't really edit it. I can't really paraphrase it. I just need to work on prompts. I can use what was generated before. I cannot really merge it. So I thought like I'll give one button contextual compose and copy kind of thing. So when I actually click on click compose, it gives you to an editor and Gmail itself. So you can again paraphrase, use synonyms, thesaurus and you can just work in your limited space to get your output. And we are just testing this for premium users. It will be rolled out for freemium just I think in a couple of weeks, but I think we are just playing with it and we are just seeing some improvements. Third is adaptive AI design and here is the irony. We want to design AI with AI, right? And I know there will be many talks on this, like how we leverage these tools. There are like thousands of tools available. There are sites available that give you freemium as well as premium, but we had quite a bit really tight to use this. So for research, you just charge up it as your buddy, but see if it is more contextual. For a brainstorm, we are starting using Jambot, which is a very cool tool to really ideate and innovate. And there are some other tools as well. And for design, we basically use mid journeys or delis to just, I have visual concepts. They may not be perfect in terms of what user message they are trying to give, but they will do the job. So for UX, I want to skip it standard for an AI project when I'm doing my MVP. Just make it functional and non-intrusive. Don't add jazz, don't add animations. This is not what they came for. You are just adding an extra value in this AI. Just keep it very straightforward. Have standard UX, use design systems, and try to use natural copy. Just don't sound robotic. Whenever I'm talking to chargeivity, I feel like I'm talking to Vicky from Small Wonder if you have ever listened to it. So coming to principles now. So when we are working this ecosystem and we were asked to generate principles because we have so many projects. And I just went to chargeivity and say, create some principles for an AI language company. And it did. The hit had no context. It just gave me 13 principles. I gave the same prompt, it gave me more principles. So if I look at it, I can't really use this data. How can I evolve this? So I broke down what are design principles and I got it that they are essential and specific value statements. I can go into deep, but for now, just consider that these are value statements that are used to define design. It may be different for all different companies. Then again, I went to chargeivity and said, give me what are the values of design? So it gave me some 200 values out of which we reduced to 40. Then I entered this conflicts in my teams because how do I know which values which team likes? So for my product managers, I gave them the same set of 40 values and asked them to like five and dislike five. I gave my design the same value. And what really happened it? My designers really liked accessibility and user-centered design, but they hated velocity. And my product managers liked velocity, but they hated the uniqueness or consistency. So how do I combine them together? How do I combine all these teams to work in a seamless flow? So again, evolving, brainstorming, we just created a first draft of what can be our principles. No need to go into deep because there will be a difference for you, but I'll just give you a couple of examples. So when you're adapting with AI, leverage AI as much as possible. Try to just use best U.S. practices and don't reinvent the wheel and just use try to use natural copy. Lastly, the couple of steps are pretty straightforward to be using design thinking a lot, iterate and validate, where we created design principles. I just focused on a couple of them which is simplify to AI and the need for speed. If I talk about simplify to AI, all our products, we just want to simplify the features and through AI. We can't have anything complex. That's a straightforward principle for us. If I look at design decisions and trade-offs, it has to be strategic, high-impact, easy to use and discoverable. So if you are working on an initiative which is strategic and high-impact, just tag this principle and everybody in the company will know that this is serious, yeah? But if you're going for need for speed, everybody will know that you can skip boundaries, you can hop, you can jump. Because it is tactical, it is low-impact, easy to build and functional. So you may leverage your intuition into it, right? So if you look at need for speed and simplify, they're clearly conflicting, right? You can't really use it on the same projects. So we classify initiative based on our principles and we have zero to one products. You can have simplify to AI. You can even have more aspirational which is called transient writing which we are really, really trying to evolve a product by adding some really class features. But if you're using some small features, you can have need for speed, but I want to start and end clearly. I just want to define how this workflow will be. So then we also try to drink our own champagne where we will just use our products in-house itself. So we want, we are writers, sorry, we are not writers, so we want to write. So I'll ask everybody to just watch a movie and do a review and then use our tools. And then I get wonderful insights because they have been developing the product, but they have not been using it. So I get really, really good insights. Also, after you have done with your project, you need to have a good adoption strategy. So think of premium first. You can introduce with a larger TM and then you introduce to the bigger audiences. Have contextual nudges as much as possible. So whenever they are triggering this flow, then have the nudges. And if you have time to onboarding, but I really don't like onboarding because it is on your face and people tend to skip it. But it's a good exercise where you can just build something very quickly in just a bunch of slides. So when you're doing iterate and validate, you do A-B test. The first initial things on gending principles, we did a lot of A-B test with the users as well as internal stakeholders. Try to be the user as much as possible. So it will really help you get in the shoes and see if you can use the product and you have a clear sort of adoption strategy. So I'd like to conclude with learn and evolve. I think pretty straightforward, just use analytics and use a feedback. We do use amplitude, we use surveys, all our product have feedback loops. It really help you to evolve your product. So summarizing scale is evolving tech by AI-driven design. You see the evolution. You try to capture the constraints. You have adaptive AI design, you iterate and validate, and you learn and evolve, right? So this is a framework where we're using for like past three months and we continue to test this in the next quarter, okay? And yeah, I hope to see some good results. And if you try to see and use into your framework, just reach out to me. I would like to take the community to get feedback if this is really working for you. To conclude, I just want to thank you and we just do really sit in my window. And it runs in the family. My kid is also chasing a pigeon. So thanks.