Automated Salesperson

Building an Automated Salesperson was an interesting project. The customer wanted an AI Salesperson that could make cold calls to an extensive list of customers that had not done business with them in 2 years or more. This system used a custom training model to bring a branded engagement to each customer. Due to this list being prior customers, the type of engagement was simple and straight forward. The model was given a very small list of directives and goals.

Tools Used to make an Automated Salesperson

Tailored GPT-4 Model

Python Programming Language

MySQL Database

Notable Mention

This project used a custom voice model to bring a unique experience to the cold call.

Why The Customer Was Interested In Trying This Method

The customer stated that she was unable to delegate this task to her current staff and did not have the resources to hire new staff to complete this task. As a result, this was an easy decision for her as there were no upfront costs to implement an automated set of salespeople, other than providing 30 VoIP lines. This implementation was paid for based on results.

Established Directives and Goals for the Automated Salesperson

The customer’s goals were fairly simple and straightforward as the call list was made up of previous customers. This is an outline of the AI Salespersons Objectives:

1. Confirm Contact Information
2. Build Rapport through Q&A about customer’s business.
3. From Products and Services List and customer rapport data, attempt to establish an appointment for (Value Proposition 1)
4. IF FAIL: Attempt next (Value Proposition 2)
5. IF FAIL: Disengage and add notes to CRM
6. IF SUCCESS: add notes and Schedule appointment in CRM

Result of the Automated Salesperson

Phone Calls Progress Sheet

Profit to Implementation Ratio

6:1

What Went Well

1. Although the goals were limited, a 7% Conversion on cold calls exceeds expectations. It must be stated that these were not exactly “Cold” cold calls, as the calls were to people that had once done business in the past.
2. For a small period of time, a 30 “person” call center spun up and executed with great efficiency.
3. 27% of people were able to get to Goal 3.

What Could Be Improved

1. Creating a better approach to Goal 2 (Building Rapport) might increase the advancement to Goal 3.
2. the short pause between the callee finishing their thought and the caller responding could improve the overall process. This could be completed by start the text-to-voice service while the response is still being generated.
3. Some calls went excessively long in the Goal 2 Section of the conversation. This is due a lack of programming in directives to direct the callee to the Goal 3 in an efficient manner. This caused some calls to cost more in resources and did not necessarily result in a successful conversion. These long conversations could be reduced by trying to direct the conversation towards the end-goal better.

Do you have a similar list you would like an AI Salesperson to try out? Contact us today to learn more.