Overview

This proposal outlines the development of an “AI Infrastructure”, designed to streamline incoming call management for service based business owners.

The infrastructure will handle different protocols including handling frequent queries about billing, new business, payment processing, and common business questions.

The aim is to enhance efficiency and customer service, to support continuous company growth, and to convert leads into sales on autopilot.

Goals

1. Develop a user-friendly voice assistant that takes incoming calls.

2. The assistant should have the ability to:

   1. Answer caller questions based on provided business information

   2. Provide the caller with instructions regarding frequently asked questions including calls about billing/invoices, new business inquiries, and how to pay rent or common charges

   3. Collect caller information including their name, contact number, email, and reason for calling

3. Functionality Requirements:

   1. Live transfer emergency calls to the team for additional help

Packages

Our package offerings are designed to cater to different needs and budget levels while providing progressively higher levels of performance.

Historically, our Good package solutions achieve 70-80% reliability, the Better package solutions reach 80-90%, and our Best package, with the most rigorous testing and evaluation, consistently achieves 95-99% reliability.

This data reflects our commitment to continuous improvement and ensuring that our voice assistant meets your specific business needs.

1. AI Infrastructure Development (2 weeks)

   1. Development of the voice assistant using Retell and Make intended to handle incoming calls about billing, new business, payment processing, and common business questions. The voice assistant will also have the functionality to forward applicable calls to team members.

   2. A light prompt engineering evaluation process will be applied to a small series of prompt candidates to validate the best-performing prompt.

2. Good - Defensive User Testing (2 weeks)

   1. User Testing

      1. Test with a limited user base to collect usage pattern data.

      2. Review user voice calls manually to collect feedback, prioritize improvements and implementation.

      3. Another round of internal testing after new changes have been implemented.

   2. Historically, agents developed under this package have achieved a reliability rate between 70-80%. This performance level reflects our standard testing and iterative improvements during Defensive User Testing, ensuring a cost-effective solution with satisfactory call handling capabilities.

3. Better - Offensive Red Team Testing (2 weeks)

   1. Extensive Engineer-Led Testing

      1. Create a variety of new test cases based on data collected during user testing. These will be entirely new corner cases we derive from user testing data and engineer-led insights.

      2. Develop a new best-performing prompt based on improved test case handling.

      3. Another round of internal testing after new changes have been implemented.

   2. Solutions developed under this package have consistently delivered reliability between 80-90%. The additional engineer-led testing and prompt enhancements help capture more complex scenarios and improve overall accuracy in handling incoming calls.

4. Best - Eval System Led Testing (4 weeks)

   1. Custom Eval Testing Framework

      1. Create a custom success metrics for primary call objectives.

      2. Test new prompt variations against the previous best prompt as a baseline.

      3. Establish and move forward with the best-performing prompt based on a custom success metric.

      4. Complete another round of internal testing after new changes have been implemented to check for potential regressions.

   2. Our most advanced package, the Best package, has historically achieved a reliability rate between 95-99%. This high performance is a result of our comprehensive Eval Testing Framework, ensuring the utmost precision and dependability in call handling and prompt accuracy.

Period of Performance

The project will commence upon the signing of the final contract. The entire development and deployment process is expected to be completed within 4-10 weeks.

Potential Engagement Resources

• Project Manager: Overall project coordination.

• Technical Lead: Integration and API handling.

• Prompt Engineer: Voice Agent Performance handling.

Fee Schedule

Total project cost is estimated at $4,000-$15,000 (depending on which setup package is selected), broken down as follows:

AI Infrastructure Development & Good - Defensive User Testing (4 Weeks)

Project Manager, Technical Lead, Prompt Engineer

$4,000

70-80%

Better - Offensive Red Team Testing (Additional 2 Weeks)

Project Manager, Technical Lead, Prompt Engineer

$9,000

80-90%

Best - Eval System Led Testing (Additional 2 weeks)

Project Manager, Technical Lead, Prompt Engineer

$15,000

95-99%

Payments are broken down by project package detailed below:

• Starting this performance period a deposit that’s 50% of the project total will be deposited (based on which package you select).

• When the project is considered completed by Scollar IO the remaining 50% will be deposited.

• Recurring operational costs (as detailed in the “Recurring Operational Costs” section) will be billed directly from the services according to actual usage and service subscriptions.

Recurring Operational Costs

In addition to the fixed project fees outlined, please note that the operation of your voice assistant will incur the following recurring costs:

• Dedicated Phone Number: $2 per month (per voice agent phone number)

• Retell AI Usage: $1 per minute (usage-based fee)

• Make.com Operations: Approximately $10 per month (estimated usage-based fee)

• OpenAI Usage: Approximately $10 per month (estimated usage-based fee)

• OpenPhone (Optional): $33 per month (Phone system/VOIP platform if needed)

Note: These costs will be billed separately from the project’s fixed fees and are subject to variation based on actual usage and any adjustments in service plans.

Completion Criteria

Scollar IO shall have fulfilled its obligations when any one of the following first occurs:

• Scollar IO accomplishes the activities described within this SOW when delivery of the outcomes listed in the sections entitled “Goals” and “Packages” is accepted by the client.

• Both parties have the right to cancel this agreement by providing 14 days written notice to the other party.