Ali Mrani Alaoui·Portfolio

AI Rapid Fire — Sales Objection Drills with an AI Coach

An AI tutor that drills sales reps on customer objections, scores every answer in under a second, and ships a personalized performance report.

AI TutorScenario-Based RoleplayVoice + Chat Input
Industry
Sales Training & Coaching
Timeline
6 weeks
Tools
Custom code, HeyGen, Synthesia, ElevenLabs, Anthropic Claude
Learners take part in a rapid fire drill with an AI.
The AI presents a customer scenario with specific objections.
Learners explain how they would handle the situation.
They can type their answer in the chat.
...or click a suggested response.
...or speak their answer out loud.
The AI evaluates their response in under 1 second.
Learners see their performance report with strengths and focus areas.

Project Overview

A B2B SaaS sales org that had been onboarding new reps with shadowing and a PowerPoint deck of objection scripts. The VP of Sales wanted reps to practice live objections before they ever hit the phones, without taking a senior AE off their desk for an hour a week.

The brief was a self-serve practice tool that an AE manager could ship to a 200-person sales org without onboarding overhead. It had to feel like a real prospect, score every answer against the team's rubric, and produce reports that survive a Monday morning pipeline review.

Each session opens with the AI presenting a customer scenario with a specific objection and tone. The learner responds by typing, clicking a suggested reply, or speaking out loud. The AI evaluates each answer in under a second and pushes the conversation forward. Sessions close with a personalized performance report that surfaces strengths, focus areas, and a topic-level breakdown.

Under the hood, each persona carries a backstory, an objection pattern, and three escalation paths. Evaluation runs on a structured rubric co-authored with senior AEs so the scoring matches what coaches already use in 1:1 reviews. The same engine has since been adapted for customer success and a discovery-call variant for new SDRs.

What's Inside

12 customer personas

Each with a named backstory, industry, objection pattern, and three escalation paths.

3 input modalities

Typed chat, click-a-suggested-response, or voice with streaming Whisper transcription.

Sub-1 second AI evaluation

Claude Haiku scores every answer against the rubric and routes the next AI turn before the score finishes serializing.

Performance report per session

Overall score, strengths, focus areas, and a topic breakdown across opening, objection handling, value framing, and close.

Manager dashboard

Team rollups, per-rep history, and a transcript viewer for 1:1 coaching.

Train-the-trainer kit

Rubric documentation, persona authoring guide, and an 18-minute Loom walkthrough.

What the Client Provided

  • PowerPoint deck of the team's top 30 customer objections
  • 14 hours of recorded customer calls with consent
  • Existing scoring rubric from the sales coaching team (Google Doc)
  • List of the 12 customer personas the team sells against
  • Two Loom walkthroughs from senior AEs handling the hardest objections

Design and Build Process

01

Persona and Objection Library

Interviewed senior AEs and recorded the objections they actually hear. Translated each into a persona card with a backstory, an objection pattern, three escalation paths, and a do-not-do-this list.

02

AI Tutor Prompting

Designed system prompts and the scoring rubric so the AI stays in character and never breaks role to compliment the learner. Wrote a small DSL to keep persona authoring fast and consistent across the 12 cards.

03

Multi-Modal Input UX

Designed type, click, and voice paths into a single chat interface. Streaming Whisper transcription means evaluation starts before the learner finishes speaking, so the AI feels responsive even on slow connections.

04

Performance Report

Designed a one-screen report: overall score, strengths and focus areas, and a topic breakdown across opening, objection handling, value framing, and close. Coaches see the same shape across every rep.

05

Pilot, Iteration, Manager Rollout

Ran a 40-rep pilot, cut three personas that confused the rubric, then rolled out to the full 200-person org with a train-the-trainer session for managers.

Tools and Stack

AI Engine
OpenAI and Anthropic Claude for in-character dialogue and rubric-based scoring.
AI Voice
ElevenLabs for persona voice playback, OpenAI Whisper for streaming learner input.
Custom code
Custom-built drill interface, lesson structure and gating, and the manager analytics dashboard.
Delivery
Embedded directly inside the LMS, with SCORM 2004 + xAPI packages for Cornerstone, Workday Learning, and TalentLMS.

Deliverables

  • Live AI drill (web app) with team and per-rep dashboards
  • 12 versioned persona cards with scoring rubric
  • Manager analytics dashboard with team and per-rep views
  • Train-the-trainer guide (PDF) plus an 18-minute Loom walkthrough
  • Embedded directly inside the LMS, with SCORM 2004 + xAPI packages for Cornerstone, Workday Learning, and TalentLMS
  • 30-day post-launch optimization sprint (rubric tuning, persona additions)