All work

Voice Roleplay Trainer

Insurance Sales Training · Skills Practice / Voice AI

AI voice roleplay that makes skills practice measurable

A US insurance sales training firm teaches agents a six-step method for closing on the first call. Datics built the voice roleplay trainer its agents now practise with: a live call with an AI prospect, scored against the firm's own method, with a recording, transcript and specific coaching after every attempt. It runs on the web and as a desktop app for Windows and macOS.

Status
In production, ongoing
Engagement
June 2025 to present, two-week sprints
Skills practiceVoice AIWorkforce learning

Skills Practice / Voice AI

Live Roleplay

Where the agent holds the voice call with the AI prospect.

  • Scenario, personality and difficulty
  • Live voice roleplay
  • Scorecard and coaching
  • Progress and admin analytics

Firm-managed training environment; scores are advisory and reviewed by managers.

Simplified view of what Datics built

The starting point

The firm's training taught a structured, six-step sales conversation, and its agents already worked inside the firm's own tools. Practice depended on live roleplay with a manager or a colleague playing the prospect.

What had to change

Coaching depended on managers being available, feedback varied from one coach to the next, and text-based practice did not feel like a real call. Agents finished training without enough repetitions to make the method automatic.

What couldn't break

Practice had to happen by voice, the way the real call does, inside the tools agents already use, and every session had to be scored the same way against the firm's own method and questionnaire rather than a generic rubric.

Why this was hard

A believable prospect has to hold a personality, mood and objections for a whole call and react to what the agent actually says, in real time, with low latency and clean audio. The session then has to be scored against a specific multi-step method, with each objection found and timestamped.

What Datics changed

Datics built a stateful roleplay engine with prospect personalities, moods and objection styles; real-time two-way voice with noise suppression; scoring against the firm's six-step method and its 15-point questionnaire; objection analysis with timestamps; and an admin side for scenarios, scripts, users and analytics.

Shipped as an MVP inside the firm's existing extension, so agents needed no new sign-up, then extended sprint by sprint: dynamic script selection per scenario, gamification, a desktop app for Windows and macOS, and staging-to-production release pipelines with an approval gate.

Scenario, personality and difficulty

Agents choose what they practise; trainers manage scenarios and upload scripts.

Live voice roleplay

Real-time voice that adapts to the agent's pace, hesitation and departures from the script.

Scorecard and coaching

AI summary, recording, transcript and a scorecard against the firm's method, with timestamped objection analysis.

Progress and admin analytics

Levels, badges, streaks and leaderboards; sessions, average scores, completion and hardest objections for managers.

How it works

Connects to: The firm's existing agent extension and CRM, real-time voice and transcription services, and noise suppression.

  1. 1

    Set Up the Call

    Where the agent picks a scenario, a prospect personality and a difficulty.

  2. 2

    Live Roleplay

    Where the agent holds the voice call with the AI prospect.

  3. 3

    Score & Coach

    Where the call is scored, coached and added to the agent's progress.

Runs within: Firm-managed training environment; scores are advisory and reviewed by managers.

Engagement at a glance

Agents can practise the same controlled calls as often as they need, by voice, and every attempt is scored the same way against the firm's method. Managers see who is practising and which objections the team struggles with.

Platforms
Web, Windows and macOS
Delivery
Two-week sprints since June 2025
Team
Three engineers with BA and QA

Under the hood

Datics role
Business analysis, product design and full-stack AI engineering, with QA and release management.
Integrations
The firm's existing agent extension and CRM, real-time voice and transcription services, and noise suppression.
Deployment
Cloud-hosted web app plus signed desktop apps for Windows and macOS.
Production boundary
Firm-managed training environment; scores are advisory and reviewed by managers.

Technology

ReactFastifyOpenAI modelsLangGraphAssemblyAIKrispPostgreSQLElectron

What this unlocked

The same engine now powers Rehearsly, our skills-practice module for workforce-learning, CE and bootcamp platforms.

Related work

Learning-agency directory

EdTech · Directory / Marketplace

A vetted, subscription directory that helps education buyers find and compare learning agencies

A members-only directory of vetted education agencies with search, filters, shortlists, moderated comments and Stripe subscriptions, built end to end in six milestones.

EdTechMarketplaceMembership
Read case study

Leadership-development practice app

EdTech · Learning App / Mobile & Desktop

A leadership-development practice app taken from the web to iOS, Android and desktop

A daily leadership practice built on seven inner archetypes, taken from a web app to native iOS, Android, Windows, macOS and Linux apps from one codebase.

EdTechLearning appCross-platform
Read case study

Document Intake Assistant

Financial Advisory · Document Intelligence / Intake Automation

Documents read into the record and multi-step forms pre-filled, with access kept separate

An assistant that reads the decks, notes and documents fund managers submit, pre-fills multi-step onboarding and due-diligence forms, scores them and keeps each user type to its own data.

Document AIIntake automationAccess control
Read case study

Have a product with the same kind of constraints?

Tell us what already exists and what you want it to do. We will tell you what it takes.

Book a 30-minute product review