Introduction
When I rolled out my first CPQ platform, one of the primary objectives was to implement customer-focused Guided Selling. This approach differs fundamentally from traditional guided selling.
Traditional guided selling helps users find a product they already have in mind. Customer-focused guided selling, by contrast, is designed to identify the product that best solves the customer’s problem—even if the customer doesn’t yet know what that product is.
A simple example illustrates the difference. If a manufacturing company needs a cabinet to house servers in a harsh environment, a traditional configurator might ask the user to select from a list of cabinet models. A customer-focused guided selling experience instead asks about environmental conditions and operational requirements. Based on those inputs, the system may recommend cabinets with NEMA 4 ratings, offering protection against wind-blown dust, rain, sleet, snow, splashing water, and hose-directed cleaning—while remaining undamaged by external ice formation. This makes the recommendation suitable for demanding industrial environments such as factories or car washes.
The value is clear: the configurator solves a problem, not just completes a selection.
The Real Challenge: Designing Guided Selling Logic
In CPQ implementations, the real challenge is rarely the technology—it is defining the right guided selling questions and decision logic.
In my first CPQ role over 15 years ago, this process was entirely manual. I worked through extensive datasheets, marketing materials, and interviews with product managers, then translated that knowledge into over 100 pages of guided selling logic. While effective, it was time-consuming, difficult to maintain, and slow to adapt as products evolved.
Applying AI to Guided Selling Design
With the emergence of modern AI, I wanted to test whether this historically manual process could be significantly accelerated—without sacrificing quality.
Specifically, I explored whether AI could:
- Derive guided selling questions directly from product documentation
- Recommend suitable products based on structured training data
- Reduce the time and effort required to design and maintain guided selling logic
To test this, I built a simple AI-driven configurator from the ground up.
Building the AI-Driven Configurator: A Practical Example
I used a personal but realistic scenario: selecting a motorcycle for 2026.
Step 1: Curating High-Quality Training Data
I gathered 100+ motorcycle datasheets across multiple brands and models. All datasheets followed a consistent format, which is critical—AI outcomes are only as good as the training data provided.
Step 2: Data Preparation and Cleanup
Before analysis, I reviewed and cleaned the content to ensure:
- Information was well structured
- Text was machine-readable
- Irrelevant or noisy data was minimized
This step is often overlooked, but it directly impacts the accuracy of AI-generated insights.
Step 3: AI-Driven Analysis and Question Design
I then used AI to analyze the entire dataset and generate guided selling questions based on the underlying product characteristics. To avoid overwhelming users, I deliberately constrained the number of questions, forcing the system to focus on the most meaningful decision criteria.
The AI performed the analysis and surfaced the following core dimensions:
- Riding experience (beginner, intermediate, advanced)
- Primary use (commuting, touring, off-road, sport, mixed)
- Budget range
- Rider height and comfort preference
- Preference for comfort versus performance
- Maintenance tolerance
These questions closely mirror how experienced sales engineers and product specialists guide customers—without requiring months of manual analysis.
The Role of a Guided Selling Framework
In addition to product data, I also created a Guided Selling Framework document. This framework acts as a critical reference for the AI, defining how to interpret product attributes and map them to customer needs.
While the datasheets provide raw knowledge, the framework provides context and intent—ensuring recommendations align with real-world use cases, not just technical specifications.
Guided Selling Framework: How to Choose the Right Motorcycle
This framework summarizes how riders can select the most suitable motorcycle by aligning their personal profile with key technical characteristics commonly found in manufacturer datasheets. It is designed for guided selling, whether used by a salesperson or an AI assistant.
1. Key Rider Decision Factors
Rider Experience Level: Beginner riders benefit from lighter motorcycles with manageable power delivery and lower seat heights. Intermediate and advanced riders can consider higher displacement engines, more aggressive ergonomics, and advanced electronics.
Intended Use: Daily commuting prioritizes comfort, fuel efficiency, and ease of handling. Touring favors wind protection, luggage capacity, and long-range comfort. Sport riding emphasizes performance and handling, while off-road and adventure riding require suspension travel, ground clearance, and durability.
Budget Range: Initial purchase price should be considered alongside ownership costs such as insurance, fuel consumption, maintenance intervals, and availability of spare parts.
Rider Height and Inseam: Seat height and motorcycle weight directly impact confidence and control. Riders should be able to place at least one foot flat on the ground for stability, especially at low speeds.
Comfort vs. Performance Preference: Upright ergonomics and relaxed engine tuning favor comfort, while aggressive riding positions and high-revving engines favor performance. Datasheets often reveal this through geometry, horsepower, and torque figures.
Maintenance Tolerance: Lower-maintenance motorcycles typically feature longer service intervals, simpler engine designs, and proven reliability. High-performance models may require more frequent servicing and higher running costs.
2. Motorcycle Categories and Typical Fit
Cruiser: Low seat height, relaxed ergonomics, and strong low-end torque. Ideal for relaxed riding and shorter to medium distances. Heavier weight but confidence-inspiring for many riders.
Sport: High performance, aggressive riding position, and powerful engines. Best suited for experienced riders seeking speed, handling, and track-oriented performance.
Adventure: Versatile motorcycles designed for long-distance travel on mixed terrain. Tall seat heights, long suspension travel, and strong mid-range power characterize this category.
Touring: Built for comfort over long distances with large fuel tanks, wind protection, and luggage options. Heavier and more expensive, but unmatched for highway comfort.
Standard / Naked: Balanced and versatile motorcycles with upright ergonomics. Suitable for a wide range of riders and uses, making them an excellent all-round choice.
Dual-Sport: Lightweight and capable both on-road and off-road. Simple, durable designs with higher ground clearance, ideal for riders prioritizing versatility and ease of maintenance.
3. Guided Selling Recommendation Logic
A guided selling flow should ask structured questions covering experience, use case, physical fit, and budget. Based on the answers, the system narrows motorcycle categories, then refines recommendations using datasheet attributes such as engine displacement, seat height, wet weight, and service intervals. The final output should include 3–5 clear justifications explaining why the recommended category or model fits the rider’s needs.
Bringing AI and CPQ Together: From Training to Guided Selling Execution
With the training content and Guided Selling Framework in place, the next step was to operationalize the solution. I trained an AI agent using Chatbase, feeding it both the curated product datasheets and the framework document that defines how customer needs should be interpreted and translated into recommendations.
Once the agent was trained, I provided explicit behavioral instructions—similar to how one would design guided selling logic in a traditional CPQ platform, but expressed in natural language rather than rigid rule trees. These instructions defined both how questions should be asked and when recommendations should be made.
The result was an AI-driven guided selling assistant that behaves with the discipline and structure of a well-designed CPQ configurator.
Defining the Guided Selling Logic for the AI Agent
The agent was instructed to operate as a guided selling assistant for motorcycle selection, with the following principles:
- Ask questions step by step to understand the user’s needs
- Avoid making any recommendation until all key criteria are collected
- Follow a predefined question sequence to ensure consistency and completeness
The guided selling flow was intentionally constrained to six high-impact questions:
- Riding experience (beginner, intermediate, advanced)
- Primary use (commuting, touring, off-road, sport, mixed)
- Budget range
- Rider height and comfort preference
- Preference for comfort versus performance
- Maintenance tolerance
Only after gathering all responses does the agent generate an outcome. At that point, it:
- Recommends a specific motorcycle brand, model, and name from the training dataset
- Provides three to five clear justification points explaining why the recommendation fits the user’s needs
- Concludes with a professional handoff for follow-up questions
This mirrors best practices in enterprise CPQ: structured discovery first, recommendation second.
From Design to Usage: A Working AI-Driven Configurator
With this logic in place, the chatbot was immediately usable. When tested, the experience is intuitive and disciplined—the agent collects user inputs, processes them against the trained data, and produces a coherent, defensible recommendation.
What stood out most was how naturally the AI replicated the decision flow traditionally built through extensive rule modeling. Instead of months of manual logic design, the system was able to derive structure directly from well-prepared data and clear intent.
Key Learnings: The Value of AI in CPQ
This exercise reinforced several important lessons about how AI can elevate CPQ across its full lifecycle—from initial design through daily usage.
1. Faster, Smarter Configuration by Design
AI can analyze historical configurations, product rules, and successful outcomes to optimize CPQ models during the design phase. This reduces over-engineering, minimizes conflicting rules, and ensures configurators reflect how products are actually sold—not just how they are engineered.
2. Guided Selection Based on Customer Need, Not Catalog Complexity
AI-powered guided selling shifts the focus from navigating product catalogs to understanding customer intent. By grounding selection in use cases and constraints, configurators ask fewer but more meaningful questions, leading to higher confidence decisions and better customer outcomes.
3. Reduced Errors and Rework Across the Quote Lifecycle
By validating selections against technical, commercial, and compliance constraints in real time, AI significantly reduces invalid configurations. This lowers engineering rework, shortens approval cycles, and improves downstream order accuracy.
4. Higher Win Rates Through Personalized Recommendations
AI learns from historical deals, customer profiles, and usage patterns to recommend optimal products, bundles, and alternatives. The result is stronger solution fit, higher average deal values, and improved close rates—without relying solely on seller intuition.
5. Continuous Improvement After Go-Live
Unlike static rule-based systems, AI continues learning after deployment. By monitoring how users interact with the configurator—where they hesitate, abandon, or override recommendations—the system can continuously refine guided selling logic and stay aligned with evolving market needs.
Why This Matters for Modern CPQ
This approach demonstrates how AI can complement—not replace—core CPQ principles. When combined with strong data discipline and a clear guided selling framework, AI becomes a powerful accelerator for building smarter, more adaptive configurators that scale with the business.
For organizations modernizing CPQ, the opportunity is no longer theoretical. AI can be applied today to reduce complexity, improve customer experience, and drive measurable commercial outcomes.