PNSS AI Sales & Support Assistant

A concept prototype for AI-assisted sales and support workflows with grounded answers, human review and clear escalation.

Comment

Inspiration

Small and medium-sized businesses often store important information across product brochures, technical manuals, quotation files, emails and employees’ personal experience. When a customer sends an enquiry, the sales or support team may need considerable time to search for information before replying.

We wanted to build more than a general chatbot. Our goal is to create an AI Sales Engineer and Support Assistant that understands company-approved information, supports daily business workflows and helps employees respond more quickly and consistently.

What it does

This concept prototype demonstrates the proposed interface and workflow. The integrations and production capabilities described below represent the intended product direction and are not yet connected to live business systems.

The concept demonstrates how a future PNSS AI Sales & Support Assistant could:

  • Answer customer enquiries using information from an approved company knowledge base
  • Communicate in multiple languages
  • Ask follow-up questions to understand customer requirements
  • Recommend suitable products and solutions
  • Generate reviewable quotation drafts
  • Assist with technical troubleshooting using manuals and support records
  • Create appointments and calendar reminders
  • Prepare customer follow-up messages
  • Support communication through web chat, WhatsApp and email
  • Escalate uncertain or sensitive cases to a human employee

Important actions such as confirming prices, issuing quotations, sending emails and making appointments remain subject to human review and approval.

How we built it

The proposed production architecture uses an agentic workflow combined with retrieval-augmented generation (RAG). The current submission demonstrates the intended experience through an interactive concept prototype.

In a production implementation, company documents such as product catalogues, specifications, technical manuals, frequently asked questions and service records would be processed and indexed in a searchable knowledge base. When a question is received, the system would retrieve relevant company information and provide it to the AI model as context before generating a response.

Specialised workflow agents could subsequently handle tasks such as:

  • Product selection
  • Sales enquiry qualification
  • Quotation preparation
  • Technical troubleshooting
  • Appointment scheduling
  • Customer follow-up

The architecture is designed to support both locally hosted AI models for greater privacy and cloud services where external integrations are required.

Challenges we faced

One major design challenge was preventing the proposed assistant from behaving like a generic chatbot or inventing unsupported product information. The prototype addresses this through a proposed grounded-response workflow and human-review checkpoints for important business actions.

Other challenges included organising information from different document formats, maintaining accurate product data, supporting multilingual conversations and deciding when a request should be escalated to a human team member.

Accomplishments that we are proud of

We developed a practical AI assistant concept that connects conversational AI with real sales and technical-support workflows.

The concept demonstrates how a future system could understand customer intent, retrieve relevant knowledge, recommend the next action and prepare useful business outputs. The modular design also allows additional tools and communication channels to be introduced progressively.

What we learned

We learned that a useful business AI assistant requires more than a powerful language model. The quality of the knowledge base, workflow design, access control and human-approval process are equally important.

We also learned that agentic AI is most effective when each task has a clearly defined purpose, trusted data source and escalation path.

What's next

Our next steps are to:

  • Improve product comparison and recommendation accuracy
  • Connect live pricing and inventory information
  • Expand WhatsApp and email integration
  • Automate appointment creation and reminders
  • Improve quotation generation using company templates
  • Add response citations and audit history
  • Strengthen user permissions and data security
  • Evaluate a locally hosted AI deployment for greater privacy
  • Conduct real-world testing with sales and technical-support teams

Our long-term goal is to turn the solution into a dependable digital teammate for SMEs—helping them respond faster, preserve organisational knowledge and deliver more consistent customer service.

Built With

Share this project:

Updates