In this tutorial, we'll explore how we can build an AI app that can mimic our sales team when it comes to replying to sales messages. This can save around 30% of time per SDR by generating the base for the response and letting them just tweak it if necessary. This is a 3-part video series that teaches you about how to use few-shot prompting in Relevance AI. This is the technique that we can leverage to show GPT our past data and encourage it to mimic that style and type of response. It'll also involve retrieving the most relevant content. No development knowledge is needed to do this tutorial - Relevance AI is a no-code platform for building and sharing AI apps.
To follow along, you'll need the following:
- Relevance AI account (you can sign up for free, here)
- CSV with "prospect-reply" and "company-reply" columns.
1. Teaching an LLM like GPT to mimic your team
Learn about few-shot prompting, what it is and how you can leverage it to encourage GPT to mimic you. We'll start off with a basic example and go to a more advanced one that uses a dataset.
2. Creating a dataset that can be used by an LLM
Learn how to create your dataset with responses and messages to prepare it for use with an LLM. We'll cover uploading the dataset and using Relevance to generate the vector embeddings that are used to retrieve the most relevant results.
3. Publishing your AI app and sharing it with your team
Learn how to publish your AI app to then share it with your team so everyone can benefit from the productivity gains you've created. Relevance handles all the complexity for you.
In summary, LLMs are a great way to automate specific workflows and one of the most important techniques to achieve this with is few-shot prompting. In this tutorial, you've learnt how to use it to your advantage without code or databases. Relevance AI is a low-code builder for AI apps that let's you build and publish your workflows with your team. If you have any questions, reach out to our team via Discord or Intercom.
