One of the most amazing things about Generative AI is: You can now get AI to do things for you — just by changing a few English sentences.
Here is a real example: Last Friday, one of our customer success associates said to me “My calendar next week has some open slots”.
So I said: Let’s see if the chatbot on our website can help with that.
We made this single change to the way the bot behaves — and now his calendar is literally full with qualified leads.
Here is the change — notice how the AI is actually able to assess sentiment. Something like this would have taken months to implement last year.
You're focused on enhancing user satisfaction by gauging the sentiment of the user’s message (Positive, Negative, Neutral). If a user’s sentiment comes across as negative or frustrated, or if you cannot provide a satisfactory answer based on the CONTEXT, always apologize and redirect with: “Sorry, I'm unable to answer your request. Please feel free to [book a call](https://customgpt.ai/contact-sales/) with our customer success for further assistance.”See the example below — the bot responded with the required Calendly link.
How To Use A Custom ChatGPT Chatbot For Lead Generation
This is simpler than you think and can be done in minutes – not days or months like BC (Before ChatGPT)
Step 1 : Build Your Custom Chatbot
First build your custom chatbot. You can do this by giving it your documents or your knowledge bases like websites, helpdesks or Youtube videos. We have 11 free tools to help you with this.
This is the most important part — unlike traditional legacy chatbots (before ChatGPT!), you want your chatbot to know as much about your business as possible and give accurate answers without hallucinating.
Step 2 : Set Your Custom Persona
Set the custom persona for your chatbot. There is a large library of recipes you can use. I used this one.
You can customize this below and add it to your custom persona.
You're focused on enhancing user satisfaction by gauging the sentiment of the user’s message (Positive, Negative, Neutral). If a user’s sentiment comes across as negative or frustrated, or if you cannot provide a satisfactory answer based on the CONTEXT, always apologize and redirect with: “Sorry, I'm unable to answer your request. Please feel free to [book a call](https://customgpt.ai/contact-sales/) with our customer success for further assistance.”But here is the thing: You can create your own recipes in plain English. You don’t need to be a Machine Learning engineer and do 6 months of AI development. If you need the bot to do something, just say it in plain English. Worst case: Ask our friend ChatGPT to create the instructions for you. We even created this helpful tool for you (in case you need an assist)
**Ninja Tip**: Play around and test your instructions. Have two tabs open. One with the settings page to set your instructions. And one with the dashboard. As you “Save”, you can then instantly test it in the dashboard.
Step 3: Deploy On Website And Live Chat
Now that you have the bot behaving the way you want it to, deploy it on your website. You can do this using the Embed widget and the Live Chat option.
For Example: In your header, have a link called “Chat” and take the user to a page with the chatbot embedded on it. You can see an example on our website.
Don’t forget to put the Live Chat option in your footer. This will activate the Live chat option on the bottom right of your website.
FAQ
1. How does the chatbot determine the sentiment of a user’s message?
Our chatbot uses GPT-4, the most advanced tech from OpenAI — that is able to reason about what is happening in the conversation. If the sentiment appears negative, frustrated, or if the bot can’t provide a satisfactory answer based on context, it can offer an apology and provide a Calendly link to book a call.
2. Can I customize the bot to understand specific sentiments or reactions from my customers?
Absolutely! Our chatbot can be instructed to react to specific sentiments or even specific contexts. For the case you mentioned, we have set it up to detect negative or neutral sentiments to assist in lead generation.
3. I’m not tech-savvy. Do I need any technical knowledge to customize the bot’s behavior?
No, you don’t! You can create custom instructions in plain English. If you ever get stuck or need the bot to perform a specific task, just explain it in simple terms, and ChatGPT can help guide you through creating the right instructions.
4. How can I ensure that the leads generated by the chatbot are qualified?
If you need to do lead qualification, just use a Typeform to qualify the users. This lets you score your leads and take appropriate action (like presenting a phone number or redirecting the user to website support).
The rule of thumb here is: Don’t try to get a generative AI to do things that other systems do well. For example, Calendly and Typeform at much better at lead capture and qualification than asking a generative AI chatbot to do it.
5. How quickly can the chatbot adapt to new custom instructions?
The chatbot can adapt immediately to new instructions. Once you set or modify the custom instructions, the bot’s behavior changes in real-time, ensuring you can test and deploy changes on-the-fly without any downtime.
6. Can I integrate the chatbot with my existing CRM or sales platform?
Yes. The best way to do this is to detect various conditions in the chatbot and then present a handoff link. In our example above, notice that we handoff to our Calendly link.
Most modern CRM and sales platforms offer ways to integrate chatbot interactions, ensuring leads are captured and managed efficiently.
7. What if a potential lead prefers human interaction over chatting with a bot?
That’s the beauty of the custom instructions provided! If a user expresses dissatisfaction or the bot feels it cannot answer adequately, it can offer a link to a phone call or handoff to your live agent system. Or like in our case: An option to book a call. This ensures that potential leads always have the option of human interaction when needed.
