Growing up, my weekends weren’t spent at the mall or hanging out with friends like most high schoolers my age. I was trapped in my parents' store, surrounded by boxes, shelves, and countless products awaiting inventory. My job? Take photos of products, upload them to the website, and write descriptions that somehow convinced people to buy yet another brand of phone case or random kitchen gadget. It was tedious work—the same routine over and over. I felt like a robot, except the real robots weren’t around to help.
Between balancing schoolwork and the constant cycle of restocking and updating the online store, I was burnt out. The worst part wasn’t even restocking itself; it was the endless product uploads and descriptions. By Saturday afternoon, after hours of taking and editing product images, I couldn’t tell the difference between a toaster and a blender. Describing them? Even worse.
I finally decided there had to be a better way. I remembered reading about AI tools that could do things like generate text and images. That’s when I found Coze.com—a no-code platform that allows anyone to create an AI chatbot using natural language to prompt its system. It was like a light bulb flickered in my head. What if I could build a chatbot to handle all the boring stuff?
This could help me save time I spent taking product images, manually researching, and tying in product details into our website.
Well, this I did and will show you how. Below is an outline of what I’ll discuss on how I implemented its tools in building my AI chatbot
In short, Coze is a no-code tool platform for building simple to complex AI chatbots in minutes. It allows users to easily create a wide range of bots using text. The chatbots can be published in its in-app store, WhatsApp, Telegram, Discord, Slack, and a whole other bunch of social media.
To create your own chatbot, simply
When customers ask or search for an item on our website, they expect to see a detailed description of it. They want to have information on what they are buying, its quality, specifications, and what makes it stand out. To provide this, I have to instruct the system on the input data it should get, and output information it should return, and in what structure. To do this, I utilize the prompt feature.
A prompt is a natural language instruction that tells an AI system what to do, and Coze uses this to make it easy for anyone to describe what they intend to build.
These instructions are entered into the system under personas and prompts. In my case, I wanted my bot to take the name of a product and provide me with a detailed description and specifications of it.
I noticed there are two types of prompting while building.
Agent prompt, which tells the chatbot what to build, how to build it, and results to return to users.
End user prompt, which queries the built chatbot to return desired results
Coze allows for the optimization of initially written prompts so it can return the best results in a well-structured manner, including a well-detailed skill set my chatbot should have.
As the saying goes, pictures speak louder than words, and what better way to reinforce the item specifications than giving customers an image of the item they searched for and intend to buy? They don’t just get to read about what the products are made of or can do but get a visual representation of every detail. To add this level of detail to my chatbot, plugins came into play.
Plugins are what make most Coze bots run faster and more effectively, delivering the best results, much like how my parent’s store utilizes the services of established delivery companies to send out paid items to customers.
They are additional software components that extend the functionality of an application. Plugins provide specialized features and capabilities that the base application, in this case, my AI chatbot, does not have. To enhance my chatbot, I made use of two plugins: Stable Diffusion and Gemini AI.
Gemini AI provided my chatbot with information on inputted product items. It takes the name of the product from the user prompt, searches through its large knowledge base, and returns a response.
A stable diffusion plugin makes it easy to generate high-quality product images.
Much like how we have a warehouse for storage and retrieval of items. I needed to be able to store the product information generated for easy reference and retrieval instead of having to prompt the system every time I needed the same information.
Databases are used to store information to be retrieved at a later time. For my chatbot, I created a database by specifying the table name, name of fields, type of data to be collected, and if it required information.
Since Coze is a no-code tool, I can easily type in this field name, and it would be created as an SQL database.
To record data in the database, I simply specified a new skill in the prompt and persona section, referencing the database table name and its fields. Also, I enabled a file box that, in Coze, holds large files, such as images, to store the product images. This would come in handy when I needed to download the images to use on my parent’s online store. Check out Skill 3 in the image below.
In the preview panel, I tested my newly created chatbot and got the following results as shown in the image below. My chatbot generated a response when queried with a product name and image and stored the result in its database for easy retrieval when needed or the chat has been cleared.
By the time I had it fully functional, everything changed. Instead of spending hours with a camera taking product pictures and trying to give a description of the product based on my knowledge, I’d simply input the product name, and my AI chatbot would take care of the rest. It even handled multiple product variations.
The best part is it helped cut down our marketing and advertising costs because the AI-generated images were good enough to post right away on social media and campaign boards.
We didn’t have to spend huge amounts hiring a professional photographer to take product images for the store when we needed to run a sales campaign.
Now, what used to take me an entire weekend only took a couple of hours.
After building my chatbot utilizing Coze, here are my personal thoughts about it.