Project №5 · INNOVA · Applied AI
AI Assistant for Technical & VIP Support
A grounded assistant that turns verified company knowledge, product data, and CPQ rules into fast, traceable answers inside CRM.
Product Lead · AI Product OwnerScope
Research · AI · CRM · CPQ
The opportunity
Make company expertise available at the moment of need.
Technical and VIP service teams worked across a large body of product documentation, troubleshooting history, support conversations, and compatibility rules. Finding the right answer quickly required knowing where to look and how to reconcile multiple sources.
I led the creation of a specialized assistant using Claude and Claude Code to bring that expertise into one controlled workflow—without relying on unverified external information.
Knowledge engineering
Real company knowledge, prepared for AI.
The work began with discovery, digitization, cleaning, classification, and conversion—not with a generic chatbot prompt.
Knowledge base
Governed technical and service knowledge created across the company.
Support research
Interviews and workflow research with Technical Support and VIP Concierge teams.
Real conversations
Thousands of minutes of recorded customer and partner conversations transcribed into searchable text.
Technical corpus
Datasheets, manuals, service instructions, known errors, troubleshooting guidance, and internal documentation.
Grounded by design
Traceable answers—not plausible guesses.
External retrieval was disabled to reduce hallucination risk. The assistant answers from approved internal sources, exposes the evidence behind each response, and escalates when that evidence is insufficient.
- 01
Retrieve
Search multiple approved internal sources for evidence relevant to the employee's question.
- 02
Compare
Reconcile documentation, support history, CRM records, and compatibility rules before composing an answer.
- 03
Explain
Show the sources used, their percentage contribution, and links back to the supporting material.
- 04
Escalate
Route uncertain or contradictory cases to a domain expert instead of presenting a guess as fact.
CRM + CPQ
From a support question to the right product—and the order.
The assistant sees products, spare parts, accessories, generations, and configurations in CRM, then applies governed CPQ logic to determine what is compatible.
- 01
Identify
Recognize the unit, generation, serial context, and installed configuration.
- 02
Diagnose
Connect the reported issue with approved troubleshooting knowledge.
- 03
Configure
Apply CPQ compatibility, dependency, quantity, and configuration rules.
- 04
Recommend
Return the correct spare part or accessory with an evidence-based explanation.
- 05
Order
Continue from the recommendation into the existing CRM ordering workflow.
Validation before deployment
Validated by the people who own the knowledge and use it every day.
Front-line testing
Technical Support and VIP Concierge employees tested real daily scenarios, answer quality, and usability.
Expert validation
Technical Support, Engineering, Documentation, and VIP Concierge leaders verified accuracy, currency, interpretation, and safety.
CRM deployment
Only after cross-functional approval was the assistant embedded into the existing CRM workflow.
My contribution
End-to-end product ownership.
Product concept and use-case definition
Support and VIP Concierge workflow research
Knowledge acquisition and data structuring
Conversation transcription pipeline
Claude Code assistant development
Closed-source information architecture
Multi-source attribution design
CRM product catalog integration
CPQ compatibility integration
User testing and cross-functional validation
CRM rollout coordination
From scattered expertise to guided action.
Verified knowledge, delivered inside the workflow.