5 Unique Selling Points with our ChatGPT Chatbot

We used to list all of our features at our ChatGPT chatbot page. It was 7 pages of scrolling, and we realised there were other more important things we could use the space for, so we had to remove it.

However, we suspect there are a middle way here, where instead of listing all features, we can focus on some of the most important ones.

This article is therefor created to give you our 5 most important features, all of which very few if any other ChatGPT chatbot providers have.

1. Conversations

Our chatbot can hold conversations, similarly to how OpenAI's ChatGPT can. This allows you to ask follow up questions, the same way a natural conversation between two human beings evolves. This sounds easy but I can assure you it is everything but easy.

The primary obstacle you'll encounter as you start implementing this is OpenAI's maximum token count. One token is equivalent to 4 characters in English, and invocations towards OpenAI's API cannot use more than 8,192 tokens for GPT4. This implies you can only use 32,768 characters in total.

32,768 characters might sound like a lot, but the problem is that this threshold includes both all questions and all answers - And when you create a chatbot such as ours, you will need to pass in business content to OpenAI as you're having it answer questions. Typically such business content is found using VSS or Vector Semantic Search towards some internal database in your internal system. If you pass in too little content ChatGPT will not be able to find the answer to your user's questions. If you pass in too much business content, you will not have any room to spare for conversations.

Then comes the problem that your needs might not necessarily be the same as the needs of others. For instance, if you are creating a content production AI Expert System, you might want to have ChatGPT use more tokens for its answers, a lot of tokens for its training snippets, and less tokens for its conversations.

The way we have solved this is by allowing you to configure on a per model basis how much training data to pass in with each question. Below is a screenshot of the configuration of our own chatbot in these regards.

Configuring your OpenAI tokens

As you can see from the above screenshot, we want to add as much training data as possible when invoking OpenAI. In fact, we're adding 12,000 characters of training data as we invoke OpenAI for each question - While we only allow for 1,000 character long questions, and only allow for 8,000 character responses. The rest is used for conversations and becomes 11,768 characters in total.

This results in a chatbot that cannot handle very long questions, answers very accurately, but won't give you very long responses - Yet still allows you to have high quality conversations - Due to the size of the "Max Message Tokens". However, if you ask our chatbot to write an article with 5,000 words for you, it will fail or stop in the middle of the article.

This is a trade off, but for a website chatbot such as ours, configured to provide basic support and help us to sell, it's probably a good configuration. For an AI Expert System created to generate content, you'd probably want to increase both "Max Request Tokens" in addition to "Max Response Tokens". Which would reduce its ability to hold conversations.

2. Images

Our chatbot can display images inline into its own conversation. This was a monumental task for us to implement, and as far as we know, we've the only ChatGPT chatbot vendor out there with this ability. According to scientific studies in the E-Commerce sector, displaying 1 image doubles your sales, and displaying 2 images quadruples your sales. Extrapolating these figures up to 4 images, implies a chatbot that can display 4 images can in theory sell 16 times as much product as a chatbot that cannot display images at all.

Of course there's a diminishing effect somewhere up with 10 to 15 images, so I doubt you'll sell a trillion times more product with a million product images. However, I think it's safe to assume that you can easily sell 10x as much product if you've got solid product images, and you're able to show these images in your ChatGPT chatbot. Below is an example from one of our clients.

ChatGPT chatbot that shows images

Would you buy the above bag if the chatbot's answer was "I am sorry, but as an AI language model I cannot show you images"?

We've got tons of research into the subject, and facts are that a chatbot that can show images is simply a must if you want to use it in an E-Commerce system as a sales assistant. In addition it's also extremely beneficial in a customer service context, since often it's easier to provide customer support using images and screnshots than simply words. For these reasons your first question to your ChatGPT chatbot vendors should be as follows ...

Can your chatbot display images?

As far as we know, we're the only ChatGPT chatbot vendor that can answer yes to the above.

3. Collecting data

Our data collection module is another unique feature. It allows you to collect information from your chatbot's users. Such information can be name, email address, phone number, zip code, hair colour for that matter, and basically any data point you want. Each data point again can be configured to be added to the context, personalizing the experience, and also saved across multiple sessions.

This allows you to use your chatbot to gather super high quality leads. If you haven't already tried this with our chatbot, you can click the button in the bottom/right corner of this page, provide the chatbot with your name and email, at which point you're being automatically subscribed to our marketing newsletter.

We can also send such leads to your CRM system, send emails to your sales executives, or anything you wish really. Our own figures in these regards have demonstrated a 700% increase in leads from our own website as we implemented this - Which of course for any serious E-Commerce website is probably a fairly drastic figure ...

4. Funnel users

Another unique feature our chatbot have is its ability to display hyperlinks, inline into the conversation. This opens up an entirely new axiom for your chatbot, allowing you to use your chatbot as a navigational tool, to guide users to for instance product pages. Below you can see how we're using it ourselves by guiding users to our Calendly page.

We're obviously biased here, but we'd suggest that a chatbot without the ability to display hyperlinks isn't really worth much. This is especially true for the E-Commerce vertical, where you want to guide users to product pages, checkout pages, etc.

Needless to say of course, but this is yet another crucial feature, especially for E-Commerce

Most ChatGPT chatbot providers can display "references", implying what page the information it's using to answer your question originates from. We're the only one that can display inline hyperlinks, to completely unrelated pages ...

5. Remembers your name

Take another good look at the above screenshot with the Calendly link. Can you spot it? Yup, the chatbot is addressing me by my name. This is because as it started the conversation with me, it was gathering my personal information. Each data point can be configured to optionally become a part of the ChatGPT data context, allowing ChatGPT to address the user by its name.

This is saved for later, implying if I come back to the chatbot 3 months later, it will greet me with "Hello Thomas", and use my name in its conversations with me. This too have dramatic effect on conversions, simply because we enjoy hearing our own names. There is tons of research on this concluding with that if somebody calls you by your name, you are much more likely to buy product from them.

In the video below I go through all of the above features, demonstrating them, and explaining a little bit more about how they're implemented, the advantages, and why they're important. It's an 8 minute long video if you prefer to watch me demonstrate these features instead of trying them out for yourself.

Conclusion

The above are only 5 unique selling points with our chatbot. As I started out with, we had to remove our list of unique features from the website because it was 7 pages of scrolling. We've got hundreds of unique features we're not aware of anybody else out there having. Still we haven't even mentioned the Magic Dashboard. However, if I start on that, I'll need an additional 1,000,000 words to give it justice.

However, we would argue that the above isn't our unique selling points, it's the bare minimum you should expect from your ChatGPT chatbot provider. All of the above features are crucial for a successful AI chatbot implementation, and it's really quite sad we're the only vendor that's able to deliver these to you. This of course is because ChatGPT made everybody lose their minds due to its hype factor.

The hype is over now, and the industry is starting to mature - Hopefully providing more room for serious vendors, and less room for the "get rich fast" individuals. However, don't believe me? Great, go ask our chatbot ... 😉