Overview
Claude is a generative AI assistant that supports reasoning, content generation, knowledge retrieval and workflow automation. This course shows customer support and service operations teams how to put it to work on everyday tasks: sorting and prioritising tickets, writing consistent customer communications, producing structured tickets and summaries, and keeping knowledge articles accurate and current.
The first day builds the core skills of triage, prompting and customer communications. The second day moves into ticket automation, knowledge retrieval, repeatable workflows and responsible use, and closes with a hands-on project that joins everything into one support workflow.
Who it's for
Anyone working in customer support or service operations who is new to Claude and wants to use it effectively and safely in daily work. It is especially suited to:
- Customer service agents and team leaders
- Support managers
- Operations and service-delivery teams
- Knowledge and quality-assurance specialists
Prerequisites
- An understanding of customer support workflows
- Experience with helpdesk or ticketing systems
- Basic familiarity with enterprise knowledge bases
- No prior experience of Claude is needed
By the end of this course you will be able to
- Classify and prioritise tickets by intent and urgency with Claude, and measure how accurate the results are
- Write tone-consistent customer replies, multi-step communication flows and summaries
- Turn conversations and transcripts into structured tickets that follow your internal format
- Query and improve knowledge articles using Projects built on your own documentation
- Design repeatable prompt workflows and plan how Claude fits alongside your existing support tools
- Apply data-handling, permission and review practices so AI-assisted support stays accountable
Course outline
Times show the standard running order and can be adjusted when we tailor the course to your team. The trainer stays available during breaks to help anyone finish an exercise.
Welcome & orientation 10 min
How the two days run, the simulated support environment, and the sample tickets you will work with.
Claude in enterprise support 30 min
Where Claude fits in a support operation, what it does well and where it needs a person: reasoning, content generation, knowledge retrieval and workflow automation. A tour of the operational use cases we return to over the two days.
Try it: Sort the task types in your own queue into “good fit”, “needs review” and “keep human”.
Prompting foundations for support teams 30 min
One simple, repeatable method for every prompt in the course: GCES (Goal, Context, Expectations, Source). Add the habit of pointing Claude at the right source material and refining a draft rather than starting again.
Try it: Turn a weak prompt (“reply to this customer”) into a strong GCES prompt, then compare the two replies.
Support triage and classification 45 min
Automated categorisation strategies: defining clear categories, giving Claude examples, and asking for one answer per ticket in a fixed format. Extracting intent and urgency from the customer’s own words.
Try it: Classify ten sample tickets by category, intent and urgency, with a one-line reason for each.
Reducing manual triage workload 40 min
Turning the classification routine into something your team can reuse, and measuring it by comparing Claude’s labels with a human-labelled set. Known weak spots, such as implicit requests, strong emotion that hides the real issue and tickets with several issues, and how to instruct around them.
Try it: Upload a labelled spreadsheet of tickets, have Claude classify it, count the mismatches, then improve your instructions and run it again.
Tone-controlled customer replies 45 min
Tone-controlled message generation that matches your brand voice and the situation. Adapting messages for different customer profiles, from new customers to senior contacts to frustrated ones.
Try it: Write one reply to the same complaint for three customer profiles, then choose and refine the best.
Multi-step communication flows 40 min
Creating multi-step communication flows: acknowledgement, progress update, resolution and follow-up that read as one consistent conversation. Replying in the customer’s own language, with a review step before anything is sent.
Try it: Build a four-message sequence for a delayed-delivery case, including a second-language version that a colleague reviews.
Artifacts: reusable replies and checklists 35 min
Using Artifacts to turn a good result into something you can edit, reuse and share: a reply template library, an escalation checklist or a simple tracker.
Try it: Create a reply template library for your five most common contact reasons.
Quality assurance for AI-assisted replies 35 min
Ensuring quality assurance in AI-assisted operations: a short review checklist covering facts, tone, commitments and customer data before anything is sent, and spotting answers that are confident but wrong.
Try it: Review three Claude drafts against the checklist and correct what fails.
Day 1 practice: from ticket to reply 25 min
Everything from today on one case: classify, prioritise, draft and review.
Try it: Work a short queue of five tickets end to end, then compare approaches in pairs.
Close 10 min
Recap of Day 1 and what to bring to Day 2.
Welcome back 10 min
Questions from Day 1 and the plan for today.
Structured ticket content 45 min
Generating structured ticket content from a free-text customer message: summary, category, priority, steps already tried and next action. Aligning outputs with internal formatting standards by giving Claude your template.
Try it: Turn five customer emails into tickets that match your internal format.
Summaries from transcripts 40 min
Creating accurate summaries from chat and call transcripts for handover, escalation and records, and checking each summary against the source.
Try it: Summarise a long transcript for an escalation, then verify three of its claims against the original.
Querying knowledge bases with Claude 45 min
Querying enterprise knowledge bases with Claude: grounding answers in your own articles using a Project with knowledge files, and asking Claude to quote the passage it relied on. Where connectors can bring in content from services you already use, and how to keep that access small.
Try it: Set up a support Project with knowledge articles, ask five test questions and check each answer against the source.
Improving and creating knowledge articles 40 min
Updating outdated knowledge articles against recently resolved tickets, and creating new documentation from support data, with a clear review step before anything is published.
Try it: Update an out-of-date article, then draft a new one from three resolved tickets.
Repeatable workflows and integration 45 min
Optimising support processes using Claude and designing repeatable prompt workflows. Saving a good routine as a Skill so it runs the same way each time. How Claude sits alongside your existing helpdesk and ticketing tools: connectors, and a demonstration of how a ticketing system can trigger automatic classification.
Try it: Write down one regular support routine as a repeatable workflow and turn it into a Skill with Claude’s help.
Security, compliance and responsible use 35 min
Data handling considerations: what may be shared with Claude, and in which accounts. Permissions and access policies, including approval settings for connectors. Auditability in AI-enabled support workflows through a simple review record. Customer messages that try to give Claude instructions of their own (prompt injection).
Try it: Draft five team rules for AI-assisted support, then classify example data as safe or not safe to share.
Hands-on project: Claude in real support scenarios 75 min
Join the pieces into one workflow: building a triage-to-ticket workflow, producing customer-ready communications, and configuring a support knowledge Q&A set-up in a Project.
• Triage a batch of tickets and turn them into structured tickets.
• Draft and review customer-ready replies for three of them.
• Answer five customer questions from your knowledge Project and note any gaps in the articles.
Close: next steps & Q&A 10 min
Agree a few things to start using straight away and what to review with your team.
What you'll need
Claude
- A Claude account on a plan that includes Projects. Sharing Projects with colleagues needs a Team or Enterprise plan.
- Skills are available on paid plans and need code execution and file creation to be enabled; on Enterprise, your administrator enables this in organisation settings. Artifacts need the same setting.
- For the connector demonstration, connectors enabled by your administrator, where your organisation allows them.
Practice files are provided for every exercise. You are welcome to use your own everyday documents where your organisation's policy allows.