Zendesk AI Customer Support Expert Needed

We are looking for an experienced AI / Zendesk / Customer Support Automation specialist to help us turn our historical support data into a comprehensive knowledge base and AI-powered customer service system.

Phase 1 – Analyse Zendesk History

We have thousands of historical customer conversations in Zendesk. We want these analysed to identify and categorise:

- Common customer questions
- Technical support & PC troubleshooting
- Order and delivery enquiries
- Returns, collections & refunds
- Warranty enquiries
- Cancellations
- Damaged/missing items
- Marketplace enquiries
- Common responses from our team
- Complaints & escalations
- Situations requiring human intervention

The objective is to extract the valuable knowledge contained within years of real customer conversations and convert it into structured, reusable company knowledge.

Phase 2 – Build Our Knowledge Base

We want a properly structured internal knowledge base covering our products, policies, procedures, troubleshooting and customer service processes.

We DO NOT simply want thousands of previous responses copied into a database.

Information should be reviewed, consolidated and structured so AI can reliably understand:

Customer Question → Relevant Company Knowledge → Correct Recommended Response

Where historical answers conflict or company policy is unclear, these should be flagged for our team to review rather than AI making assumptions.

This knowledge base should become the central source of truth for both our staff and AI.

Phase 3 – AI-Assisted Zendesk Responses

We then want AI integrated into our Zendesk workflow.

Initially, AI should draft responses rather than automatically send them:

Customer Email → AI Analyses Ticket → Searches Knowledge Base → Reviews Relevant Context → Drafts Response → Agent Reviews/Edits → Send

AI should also learn our preferred customer service tone/style from approved examples.

Accuracy is extremely important. AI must be grounded in approved company knowledge and must not invent policies, warranty information, technical instructions or promises to customers.

Clear rules should determine when AI can confidently assist and when a ticket requires human escalation.

Phase 4 – Further AI & Automation

We also want you to identify other opportunities to automate our customer service operation, including:

- Automatic ticket classification/tagging
- Intent & priority detection
- Complaint/sentiment detection
- Suggested macros/responses
- Ticket summarisation
- Chargeback/dispute detection
- Urgent order/delivery detection
- Technical support routing
- Repeat-contact identification
- Automatically requesting missing information
- AI troubleshooting assistance
- Suggested follow-up questions
- Quality checking agent responses
- Reporting on common contact reasons
- Identifying recurring product faults/issues

We would eventually like to investigate extending the system to assist with Amazon, eBay and Temu customer messages, subject to available APIs/integrations.

Important

We are NOT looking for someone to simply connect ChatGPT to Zendesk.

We want someone capable of understanding our business, analysing historical data and designing a reliable, scalable AI customer service infrastructure.

We are open to Zendesk's native AI capabilities, OpenAI/LLMs, RAG, vector databases, APIs and third-party automation platforms. We want the successful applicant to recommend the best architecture rather than us dictating the technology.

Experience We're Looking For

Ideally you will have experience with:

- Zendesk & Zendesk APIs
- OpenAI / LLM integrations
- RAG & AI knowledge bases
- Vector databases / semantic search
- Customer support automation
- E-commerce customer service
- API/workflow automation
- Prompt engineering & AI evaluation
- GDPR/data privacy

Previous experience building AI customer service systems from historical ticket data is highly desirable.

When Applying

Please explain:

1. Your Zendesk experience.
2. Similar AI customer service systems you have built.
3. How you would analyse thousands of historical tickets.
4. How you would structure our knowledge base.
5. How you would integrate AI-generated drafts into Zendesk.
6. How you would minimise hallucinations/incorrect responses.
7. Your recommended technology/architecture.
8. Any Amazon/eBay/marketplace integration experience.
9. Estimated timeframe and approach.

Please provide examples of similar projects where possible.

Our Goal

We want to create an AI customer service co-pilot that helps our team respond faster and more consistently while maintaining human oversight.

The system should reduce repetitive work, preserve our accumulated company knowledge and allow customer service to scale with our sales volume.

If successful, there will likely be ongoing work maintaining, improving and expanding the system.