GPT-enhanced conversation

Supercharge your chatbot with GPT4-powered understanding of conversational context.

Sometimes we don't repeat essential context every time we ask a question (or even the first time we ask it). This can leave chatbots struggling to understand what we mean.

GPT enhancement adds to the AI features already built into Helpfruit, to help plug that gap between what your customer types in chat and what they mean.

How does GPT enhancement work?

If your customer asks a question and your chatbot doesn't have a good match in it's Q&A knowledge, it can get some AI help ⚡️

Our custom prompt provides GPT4 with:

• Your customer's question
• Up to 5 preceding questions (the conversational context)
• Some information about your business and your chatbot's role (the business context)

With this info, GPT4 may be able to enhance your customer's question. It will then ask this re-worded question, and see if it finds a better Q&A match.

If it does find a better match, it will provide the relevant answer or "did you mean" answer.

If it doesn't find a better match, Helpfruit's usual did you mean and no answer processes apply - including web search fallback, contact form or live chat, depending on your bot's configuration.

An example of GPT enhancement

A customer has asked the EVA bot "how to intergrate with teams". With GPT enhancement, the bot gets additional context that helps it to understand the question and provide a relevant answer.

Click the AI button in Q&A insights to see how AI has enhanced the conversation.

The AI enhancement to the original question is indicated in red. This is only visible in the portal - it's not something that your bot user will see. For them the experience is seamless.

Are there any risks with using GPT enhancement?

The only answers provided by your chatbot are those already created and curated by you in the Helpfruit portal. AI is just used to enhance the question. This means that your bot can't "go rogue", "hallucinate" or make things up. It can only answer with your content.

You have full visibility of the AI intervention in the Helpfruit portal - so you can monitor how GPT is enhancing the user's question, and make tweaks accordingly.

We use the OpenAI API to power this feature. This means that any data submitted in chat is NOT used to train OpenAI models or improve OpenAI’s service offering. Please refer to Open AI's API data usage policies for more information.

How do I enable GPT enhancement?

Contact us to get started. It's helpful if you provide us with a brief summary of your bot's role - eg "Bot answers HR and IT helpdesk questions for software company ACME solutions" 

We'll use this information to write a custom prompt for your bot, and will then enable the feature. GPT enhancement can be disabled on request if you decide at a later date not to use it.