AI · Learning Experience

AI-Powered
Customer
Segmentation

Designing Conversational AI Experiences to Transform Sales Training at Cox Business

Client
Cox Business
Year
2025
Platform
Surge9 AI
Role
Sr. Instructional Designer
Simulation Suite Active
9 AI Customer
Personas
27+ Industry
Conv. Paths
1.2K Customers
Surveyed
AI EngineSurge9
DisciplineConversation Design
EvaluationAI-Generated
EnvironmentOn-Demand, Scalable
Overview

Cox Business launched a customer segmentation strategy designed to help sales representatives personalize conversations based on customer needs, motivations, and buying behaviors. However, traditional training methods lacked opportunities for sellers to practice these conversations in realistic environments.

As lead AI Experience Designer and Conversation Designer, I designed and developed a suite of nine AI-powered conversational simulations using Surge9. Each simulation featured dynamic AI personas capable of replicating authentic customer interactions across multiple industries, allowing sellers to safely practice consultative selling skills in a scalable, on-demand environment.

The Problem

Sales representatives needed a better way to practice customer conversations before engaging with real clients. Existing training solutions presented several challenges:

  • Training focused heavily on knowledge transfer but offered limited opportunities for skill application.
  • Live role-play exercises were difficult to scale and often inconsistent.
  • Sellers struggled to adapt their communication styles to different customer personalities and buying behaviors.
  • Managers lacked scalable methods for providing consistent coaching and feedback.
  • The organization needed a way to reinforce customer segmentation strategy through realistic practice experiences.
Design Challenge

How might we use AI to create authentic, scalable customer conversations that improve seller confidence and performance?

Research & Discovery

This project was heavily research-driven. Findings revealed that successful sellers adapt not only their solutions, but also their communication style, pacing, technical depth, and relationship-building approach. Each customer segment demonstrated unique emotional drivers, fears, and decision criteria. These findings directly informed the AI behavior models.

Stakeholder Interviews
  • Sales leaders
  • Customer segmentation experts
  • Subject matter experts
  • Change management teams
  • Sales enablement leaders
Data Sources
  • Customer segmentation research
  • Survey data from 1,200+ customers
  • National customer datasets
  • Existing customer personas
  • Sales performance insights
Key Findings
Sellers must adapt:
  • Communication style
  • Pacing and technical depth
  • Relationship-building approach
Each segment has unique:
  • Emotional drivers and fears
  • Decision criteria
  • Buying behaviors
Solution

I designed an AI-driven learning ecosystem consisting of nine customer personas, each representing a distinct business segment. The solution moved through four interconnected design phases, culminating in a fully integrated learning experience that extended beyond the simulations themselves into how learners discovered, accessed, and revisited the content.

9 AI Customer Personas
27+ Industry Conv. Branches
100% AI-Generated Evaluations
Phase 01

AI Persona Design

Each of the nine AI personas was built as a fully realized character, not a template. I developed detailed behavioral profiles containing background stories, personality traits, emotional drivers, professional goals, pain points, buying behaviors, and conversation styles. The goal was for each persona to feel like a real person a seller might encounter in the field.

Example Persona Profile
Marcus T.
VP of Operations · Mid-Market Manufacturing
Drivers
Operational efficiency
Cost reduction
Team productivity
Fears
Downtime risk
Budget scrutiny
Vendor reliability
Style: Direct and data-driven. Skeptical of vendor claims. Responds to ROI evidence over feature lists.
Phase 02

Conversation Design

I developed conversational frameworks that enabled AI personas to behave dynamically rather than following a script. Each simulation could ask realistic follow-up questions, challenge weak recommendations, demonstrate objections, adapt its tone based on the learner's responses, and simulate authentic customer behaviors. The conversations branched naturally based on seller performance.

Sample Conversation Exchange
AI Persona (Marcus)
"Before we go further, what makes you confident your solution handles the kind of call volume we're running across three shifts?"
Seller
"Great question. Our platform scales dynamically, and we have clients in manufacturing running similar shift structures."
AI Persona (Marcus)
"I'd want to see some data on that. What does uptime look like during peak hours?"
Phase 03

Prompt Engineering

Designing the AI personas was only half the work. Getting them to behave consistently required careful, iterative prompt engineering. I authored and continuously refined prompts to ensure each AI persona stayed in character, reflected segment-specific behaviors, avoided hallucinations, and delivered consistent evaluations. Multiple iterations were tested and refined based on stakeholder feedback.

Prompt Architecture
// PERSONA CONTEXT You are: Marcus, VP of Operations Segment: Mid-Market Manufacturing // BEHAVIOR RULES Rules: Stay in character at all times Challenge weak recommendations Ask follow-up questions naturally Never acknowledge AI identity // EVALUATION OUTPUT Score: discovery, empathy, alignment
Phase 04

Evaluation Framework

Each simulation provided immediate, AI-generated performance feedback that evaluated learners across six consultative selling competencies. I developed custom scoring models that assessed not just whether sellers got the right answer, but how they engaged, how they listened, and how well they connected solutions to customer needs. This allowed for meaningful, consistent coaching at scale.

AI Evaluation Scoring
Discovery Skills
8/10
Empathy
6/10
Solution Alignment
9/10
Industry Relevance
7/10
Consultative Selling
7.5
Communication
8.5
Experience Delivery & Adoption

Learning Hub Design

To ensure learners could easily discover, access, and revisit the AI simulations, I designed a SharePoint-based learning hub that served as the central entry point for all segmentation resources. The hub housed the AI simulations, explainer videos, segment reference materials, and supporting documentation in a single, easy-to-navigate experience. The design prioritized clarity, low friction, and visual consistency with Cox Business brand standards.

Cox Sky Blue #00AAF4
Cox Business Dark Blue #002F87
White #FFFFFF
Challenges

Every phase of this project surfaced real constraints that required creative problem-solving and iterative refinement.

Challenge

AI Evaluation Accuracy

Early testing revealed inconsistencies in AI scoring and feedback. Learners occasionally received high scores despite missing critical customer needs.

Solution

Refined evaluation prompts and introduced stricter performance criteria through multiple testing cycles.

Challenge

Stakeholder Alignment

Multiple stakeholders across sales and enablement teams required ongoing feedback and sign-off throughout development.

Solution

Established recurring review sessions and iterative testing cycles to maintain alignment at every stage.

Challenge

Evolving Requirements

Industry assignments changed late in development as new customer data became available, threatening the existing architecture.

Solution

Built a modular conversation architecture that allowed industries to be swapped without rebuilding entire simulations.

Challenge

Balancing Complexity

The simulations needed to support both novice and experienced sellers without creating dead-ends or oversimplified conversations.

Solution

Designed AI conversations to scale naturally in difficulty while avoiding technical dead-ends for any skill level.

Results & ROI

The project delivered a complete, scalable AI-powered training ecosystem that the organization had never had before. Although formal business metrics are still being collected, the immediate value and organizational impact were clear.

9 AI customer personas designed and deployed
27+ industry-specific conversation paths
AI-generated performance feedback at scale
On-demand, scalable practice environment
Full performance support ecosystem
Modular architecture built for expansion
"

The next best thing to a live customer conversation.

Stakeholder Feedback · Cox Business Sales Enablement
Reduced dependence on facilitator-led role plays
Enabled consistent sales practice experiences at scale
Increased stakeholder confidence in AI-enabled learning
Sparked expansion discussions across additional business units
Expected Business Outcomes
  • Faster seller ramp time
  • Improved customer conversations
  • Increased seller confidence
  • Reduced coaching time
  • Scalable practice opportunities
  • Lower training delivery costs
Long-Term Architecture Value
  • Outside Sales expansion
  • Indirect Sales applications
  • Additional customer segments
  • Future AI coaching initiatives
Reflection

This project fundamentally changed how I think about learning design. Rather than designing content, I was designing conversations. Success depended not on information delivery, but on creating authentic interactions between humans and AI.

Skills Strengthened
AI Experience Design Prompt Engineering Conversation Design Human-AI Interaction Design Systems Thinking UX Research
If I Were to Continue Iterating
Adaptive AI difficulty levels
📊Enhanced analytics dashboards
🎙Voice-first interactions
🤖AI coaching recommendations
🔗CRM and LMS integrations
Ultimately, this project demonstrated how AI can move learning beyond content consumption and into authentic skill development through practice.
Tools & Technologies
Surge9 Microsoft Copilot Adobe Photoshop Synthesia SharePoint Microsoft Office Suite
🔒

Protecting Client Confidentiality

Throughout my career, I’ve partnered with organizations across healthcare, financial services, technology, and telecommunications to design learning experiences, digital products, and communication campaigns.

Because much of this work contains proprietary information, internal processes, or confidential business content, the portfolio examples presented here have been selectively edited. Rather than displaying complete courses or full project deliverables, I’ve included representative sections that demonstrate my design thinking, creative approach, and technical execution while honoring the confidentiality and trust of the organizations I’ve worked with.

Protecting client information is just as important as showcasing my work.

Back to Work
me + you — peace Now Playing
0:00