See what CRSTA does — and how you’d build it.
Live demos across five industries and two levels of conversational capability. Watch the demo, then watch the explainer to see the development process behind it.
What are Levels?
A short walkthrough of what CRSTBL Conversation Agent Levels mean and how to think about them for your business — before you dive into the demos below.
How to think about the Levels.
The Level of a CRSTA deployment is defined by how many context layers are required to achieve a given conversational capability — and how much design work goes into building them.
Foundation Conversations.
Direct-answer conversations built on 1–2 context layers over general business information.
- Content: website scrape + menu / catalog / offers upload
- Design: general conversation instructions
- Deployment: fast — days, not weeks
- Fits: single-location businesses, standard product/service lines
- Optional presentation: avatars or voice-only agents
Typically fits a CRSTBL Local deployment. (Min. 1 day after account activation.)
Workflow-Driven Discovery.
Multi-step conversations that ask a sequence of questions to arrive at a recommendation — grounded in a connected business database.
- Content: connected product database + structured attributes
- Design: multiple context layers + question / recommendation logic
- Deployment: consultant-led design and integration
- Fits: distributors, technical products, guided sales, service configurators
- Optional presentation: avatars or voice-only agents
Enterprise-scoped individually. Requires CRSTBL project consultant time — typically 4–8 weeks to go-live.
Foundation Capability Demos.
Five live CRSTA agents deployed on real business scenarios. Each was built with 1 context layer over standard business information.
Click any video below to enlarge →
Menu discovery for a busy dinner service.
Freddy's Cocina · A guest browses Freddy's menu, asks about dietary options, and moves toward an order — no waiter, no download, no wait.
See the industry pageVoice-driven product recommendations.
Falcon Landing · A shopper on an e-commerce site asks product questions by voice, gets recommendations, and gets routed to the buy page.
See the industry pageProduct discovery for potential customer.
UnFo · Real-time product matching, facilitate account setup.
See the industry pageService intake and lead qualification.
AA Business · A CRSTA agent deployed on a services website — capturing lead intent, answering FAQs, and routing serious inquiries.
See the industry pageHow a Level 1 agent is designed — and how it performs live.
John Chang walks through the four design elements of a Level 1 Conversation Agent (presentation, persona, personality, context) and demos a live restaurant deployment for Freddy’s Cocina — ordering, catering inquiries, multi-turn recovery, and where a single-context-file build fits.
Conversational Discovery Workflow — Level 1
The design and deployment sequence behind Foundation Capability agents.
Business Data & Knowledge
Product catalog, website content, product sheets, menus, offers.
CRSTA Knowledge Engine
Structures your business information into machine-readable context.
Design Context based on Intended Conversation
1–2 context layers scoped to the conversation types you expect.
Test and Deploy
Publish to your website; capture conversation transcripts for intent analysis.
Business Intelligence Feedback Loop
Reporting on questions, gaps, and demand patterns feeds ongoing conversation design.
Static website → conversational business intelligence.
Before CRSTA
- Static website
- Visitors search or leave
- No visibility into unanswered questions
After CRSTA
- Visitors have conversations
- Questions are answered instantly
- Every interaction becomes actionable business intelligence
Workflow-Driven Discovery Demos.
Level 2 capabilities include multi-step conversations connected to large sets of business data and verbal Q&A workflow. Anticipated contextual load and length requires more content management and design on the business’s part.
Technical Sales support and Product expert.
A distributor agent asks a sequence of questions about the customer’s application, then recommends the right glove from the catalog.
One expert adviser per loan type.
Qualify.com · A restaurant owner exploring SBA and equipment financing is guided by specialized advisor avatars — Sam for SBA loans, Eddie for equipment financing, Bianca for business loans (with multilingual support).
Multi-avatar Level 2: one expert advisor per loan type.
John Chang walks through how Qualify.com uses a distinct AI advisor for each SBA loan type — and the design tradeoffs that led to a multi-avatar Level 2 workflow instead of a single generalist agent.
Conversational Discovery Workflow — Level 2
What’s different when the agent needs to guide the customer to an answer.
Business Data + Connected Systems
Product database, CRM, inventory, order history, session state — not just static content.
CRSTA Knowledge Engine + Multi-Layer Context Design
Multiple context layers linked to structured data and workflow logic.
Design the Discovery Workflow
Question sequences, constraint gathering, recommendation logic, escalation paths.
Test and Deploy
Website, kiosk, avatar, voice — with rich intent and workflow-completion capture.
BI + Semantic Mapping + Refinement
Continuous refinement based on where customers get stuck and what they choose.
Static Q&A → guided decision-making.
Before CRSTA Level 2
- Customer describes need — sales rep interprets
- Product-fit expertise trapped in a few key people
- Long lead times to answer technical questions
After CRSTA Level 2
- Customer guided to the right product in one conversation
- Product-fit expertise available 24/7 across channels
- Sales team focuses on qualified, high-intent conversations
Want to see this working on your business?
Schedule a call with the CRSTBL team — we’ll walk through your business and show what a Level 1 or Level 2 deployment would look like.