The first stage while designing a bot should always be researching. Many people think you can start building a bot with just the essential functions you'd do with other web services; at the end of the day, all you want is for the world to meet your bot and find it helpful, but that may not be true.
‍
While you can give your bot superpowers to solve people's lives, they will only use it if it feels easy and comfortable. So in that order, you need to know your users first. Here are some tips for conducting better bot research:
‍
1. Know why: Determine why you are creating this bot. Is it for e-commerce? Is it for leads? What kind of business or project are you making it for?
2. Ask for help: Two heads think better than one, better three, four, or an entire team. Surrounding the project with people that is expert in their areas and the business is going to give fresh perspectives for your bot, making sure you cover a lot of different scenarios.
3. Understand your users: This is basic UX Design, but in this field, you could be getting insights from an active tool like conversations on social media, metrics from the business website, call center, or even interviews on a physical store.

Pro tip: Make a list of all your research needs so that you can send it to your clients, and they can start working on it before defining the MVP.

Here's the deal, if you don't define an MVP, you'll end up with a colossal diagram full of possibilities, a year-lasting roadmap, and a bot that'll see the light when your user's needs have changed. Believe me, it has happened to me, and every conversation design passionate.So let's review what you need to define an MVP that ensures the success of your bot:

1. Establish use cases: Group the results you obtained in your research by topic. Some questions that can help you create them are: Why do my users message me? What could they request from the bot? What proactive solutions can I offer them? 
2. Meet with the technology team: To design a successful bot, you must know all the functionalities you can include. This will also depend on the platform on which it will be developed (Dialogflow, Amazon Lex, IBM Watson, etc.). 
3. Prioritize: A bot that can resolve 100% of queries would be great, but it would also require endless work and unlimited resources (at least for now). Therefore, you must choose which use cases will be the most requested and start from there. Remember that some cases sound fantastic technologically, but only 5% of your users will use it. Don't lose sight of the fact that you design for them, and that sometimes (most of the time) the most effortless questions to resolve are the most consulted ones. If you need help with defining your MVP, I recommend using the Pareto principle, which suggests that 80% of the problems come from only 20% of the causes.

‍Pro tip: Detect errors in other user journeys. Suppose you're designing a customer service bot. In that case, it will be helpful to identify the opportunities that other brand products have, so when your users come looking for an answer to their problem, they'll feel that the knowledge is consistent.

The ideation process may be the most fun; at this point, you should unleash your creativity and develop solutions for the use cases you defined earlier.Below you will find a list of components that all bots should have:

1. Entry points: Where do your users come from? Are they coming from a web campaign? From a physical flyer? Are they entering organically by typing a greeting? First use and greetings: Bots usually greet in the same way on every occasion, perhaps due to the scripts used in call centers. However, we can give a warmer touch and identify the first time we talk to the user, and subsequent times, giving different types of greetings and instructions. 
2. Navigation buttons: "Back" buttons are underestimated. On many occasions, I have come across bots that contain endless dead ends, negatively affecting the user experience. Keep in mind: If your user doesn't find the answer they're looking for, they may be able to find it later or prefer to talk about something else. Allow them to navigate comfortably within the flow you designed; this way, they will know everything you can help them with, and their acceptance will be greater. 
3. Not understood: "I'm sorry, I don't have that answer" sound familiar? One of the most common mistakes when designing bots is not anticipating what will happen in case of an error. You should consider that your user is not getting a solution to their query so they will feel frustrated. I recommend not letting them fall into this point more than three times and providing another way out. There are different ways to handle this situation, for example, using a search engine that returns the most frequently asked questions closest to the user's sentence. Another way is to offer to speak with a human (Yes, bots need humans). While the intention is to reduce the number of queries that go to an agent, there will always be specific cases that require reasoning that only we can provide. If none of these is feasible for you, you can always let the user know that the bot will keep learning and give them the option to return to the main menu.

‍Pro tip: Less is more. Remember that interactions within a chat aren't always comfortable for long conversations, especially when it comes to bots. So try to design a clear, simple, and easy-to-digest experience. The general recommendation is to use positive/negative questions, button sets of no more than five options, and forms with a maximum of 5 questions.

You have already designed a conversation diagram to serve you for internal design testing. However, creating a functional prototype to test with users can give you insights you won't get any other way.

Your prototype doesn't have to be high fidelity; your testers can be members of your family, friends, or even people you just met. You may not have the same results as with a large budget, but it's guaranteed that you'll get a point of view you didn't have before. Here are some prototyping tools:

1. Pen and paper: All you need is imagination and some drawings to simulate the experience, place them in order, and depending on what the user selects, give them the next page. 
2. WhatsApp Thread Builder (wathreadbuilder.com): The official WhatsApp prototyping tool is free and enables you to create images of your bot's appearance, including buttons, multimedia elements, and e-commerce. However, it may be limiting, particularly if you're designing for platforms with more functionality. Nevertheless, it's sufficient for most use cases.
3. No-code platforms: Many platforms allow you to build your bot in hours without writing a single line of code. Many of them are free and will allow you to develop a high-fidelity prototype with which your users can interact in real-time.

‍Pro tip: Prototype specific functionalities. You can test each use case at a different time and note how often users select options not yet developed, so you'll know which one to continue with.

There are different ways to test a bot. In this case, we will look at the main two: 

1. Guided tests: These types of tests will help you detect improvements in the design. To do this, you must define different journeys so your users know their goals within the bot. Try not to limit them; if they want to take another path that is not the journey, it will work as long as they can reach the set objective.
2. Open tests: Consist of delivering the prototype to the user and asking them to interact openly. These tests will work best on a bot in production, as they will test its intelligence and ability to solve problems.

‍Pro tip: Write down everything you see and how often requests, errors, and intentions from users are repeated. You can use it to improve your design.

Have you ever thought about what makes a bot premium? Firstly, natural language processing (NLP) and other tools allow us to respond more accurately to users. However, this is already a reality in different bot development platforms. Then we have Generative Artificial Intelligence (like ChatGPT), which understands different contexts and can create content instead of repeating it.

However, this kind of technology is so new that it is currently costly, and because of the complexity it requires, it is challenging to give it a personality.

‍How do I make my bot stand out from the rest?

‍The answer is simple: Personality and Writing.

Have you noticed how Alexa can tell you jokes or how Siri has hidden jokes within certain questions? These are some of the attributes that characterize particular assistants. When we build a bot, especially for businesses, we differentiate it through the same characteristics that give visibility and essence to the brand. Writing clearly, concisely, and specifically for conversational experiences allows us to reduce the cognitive load on users and provide relevant information. Moreover, by giving these texts personality, we will have a bot that reflects the values of the brand as well as its essence, and we will generate homogeneity among its different channels.

‍Pro tip: A bot with a personality does not guarantee that it has the right personality. I recommend exploring different brand archetypes to find the one that fits the solution you want to create.