How ElevenLabs Built AI Role-Play Coaches to Train Hundreds of Sellers Globally

The Challenge

ElevenLabs was scaling its global go-to-market team rapidly. New sellers needed to master complex technical concepts and pitch them credibly to enterprise buyers. The existing enablement model relied on slide decks, recorded demos, and peer role-play, which had three fundamental problems:

Inconsistent practice quality. Peer role-play depended on partner availability and willingness to push back. Most sessions became walkthroughs, not realistic buyer conversations.

No objective measurement. Managers had no visibility into whether a rep could execute a pitch until they were in front of a real customer.

Couldn't scale globally. With reps across North America, EMEA, and APAC, coordinating live practice sessions across time zones was impractical.

The Approach

The GTM Enablement team built a series of AI role-play coaches using the ElevenAgents platform. Each coach simulates a realistic enterprise buyer persona that reps practice pitching to via live voice conversation, not text chat. Reps practice the way they actually sell: speaking, listening, and thinking on their feet.

Realistic buyer personas with hidden information.

Each AI buyer persona holds information at multiple tiers. Basic context is shared freely, but deeper details like compliance requirements, past vendor failures, or budget constraints are only revealed when the rep asks the right questions. Sharp questions earn rich detail; vague questions get flat answers.

Structured beats, scoring, and analytics.

Each role-play follows a defined beat structure that mirrors a real customer conversation, with the rep demonstrating specific skills at each stage. After the conversation, the agent breaks character and delivers coaching feedback, scoring three holistic criteria on a 1-3 scale: technical accuracy, buyer credibility, and conversation altitude. Structured data is extracted from every transcript automatically, feeding leaderboards and giving enablement leaders visibility into team-wide skill gaps without reviewing individual recordings.

On-demand teaching when reps get stuck.

At any point during the role-play, a rep can break character and ask the AI coach for help. The agent pauses the scenario, teaches the specific skill with concise guidance and sample language, then offers to resume so the rep can try again. There is no penalty for using teaching mode, turning the agent from a pure assessment tool into an active learning environment.

The Results

The program launched globally and achieved results that exceeded expectations:

Over 90% completion rate across hundreds of reps in all regions, up from 40-60% with peer role-play. The AI coaches removed the scheduling friction and inconsistency that held reps back.

Perfect 5/5 feedback score from reps on the training experience. Reps reported the AI buyer felt realistic, the feedback was actionable, and the ability to retry with teaching mode made them feel like they were improving, not just being tested.

Skill-specific insights at scale. For the first time, enablement leaders could see aggregate data on which conversation beats reps struggled with most, informing follow-up training and coaching priorities.

What We're Measuring Next

The program is still early. With the engagement and adoption data validated, the team is now focused on two deeper measures:

Key skill competencies and gaps. Using beat-level evaluation criteria and Data Collection, the team will map which specific skills reps are strongest and weakest on across the organization.

Impact on sales conversion metrics. The team will track discovery-to-trial conversion rates, win rates, and deal sizes for reps who completed the program versus baseline, connecting enablement investment directly to revenue impact.

The Future of AI Role-Play

The current program trains reps on generic buyer personas that represent common industries and use cases. The next evolution is personalized, deal-specific practice: AI coaches that simulate the actual person a rep is about to meet.

Before the call. The AI coach builds a persona from the buyer's LinkedIn profile, company history, industry vertical, and CRM data. The rep practices with a simulated version of their actual buyer, learning their likely priorities and objections before the real conversation.

After the call. The rep jumps into a post-call coaching session. The AI coach identifies where they lost momentum or missed an opportunity, then runs a targeted practice scenario focused on exactly those gaps.

This shifts AI role-play from scheduled enablement to a continuous practice layer that wraps around every deal, compounding improvement across the entire pipeline.

Try It Yourself

Experience an AI role-play coach firsthand. Click the link below to practice a sample scenario and see how the tiered disclosure, real-time coaching, and automated scoring work in a live conversation.

DEMO AGENT LINK: https://elevenlabs.io/app/talk-to?agent_id=agent_8101kpnqyvgze3c93fvgwpjtxrtj&branch_id=agtbrch_2901kpnqyx34femb78njqe6h7fhc

AI ROLE PLAY GUIDE: https://docs.google.com/document/d/1J7iEezVwBJFXXA_DJ_lekSw2CuBs3h9vlqD3pjP1IuA/edit?usp=sharing

To learn how your organization can build AI role-play coaches for sales enablement, customer success training, or any practice-based learning program, contact your ElevenLabs account team or visit elevenlabs.io.