Specialised Track · Customer Service Function

Re-engineer customer service with AI.

A two-day, hands-on programme for customer service leaders, contact-centre managers and support team leads. Apply AI to ticket triage, response drafting, knowledge assistants and sentiment analysis, then build and evaluate a working triage agent in a pre-configured lab. Every delegate leaves with a working support agent, a service prompt library and a ninety-day plan for their function.

2 days 12 modules 60%+ hands-on Codespaces lab included STOA certified
Daniel Sto
Why this programme exists
"Triage, first replies, escalation. Every support motion is an agent waiting to be built, and a person still decides what reaches the customer."
Daniel Sto · Director of Faculty, STOA Institute
Duration
2 days
Format
Live online or on site
Next open date
Private cohorts
Fee
On request
per delegate
For teams
Private cohort from 8 delegates
Schedule a TNA call
Programme overview

Built around the customer service workflow, from triage to follow-up.

This two-day programme is built for the people who run customer service, support and contact-centre teams and who now have to decide where AI belongs in the service desk. Customer service runs on written work: tickets, chats, emails, complaints, escalation notes, knowledge articles and quality reviews. Generative AI classifies, drafts and summarises that work, and an AI agent goes one step further: given a goal, it reads the ticket, applies the routing rules, drafts the first reply and hands the result to a person for approval. The decisions that follow sit with the service leader: which work is handed over, under which rules, and on what evidence the result can be trusted in front of a customer.

Day one establishes how generative AI works, takes every participant through structured prompting on their own customer enquiries, and applies AI across the service workflow: ticket triage and routing, response drafting and complaint handling, knowledge assistants grounded in policy, and sentiment and theme analysis across ticket history. Day two moves from assistants to agents. Participants design and build a ticket triage and first-reply agent on a support inbox carrying realistic ambiguity and planted edge cases, instrument it and read its traces, metrics and costs, then sit as the review board on a deployed support agent's pilot output and issue a go or no-go memo. Governance for customer-facing AI and the ninety-day plan close the programme.

Every lab runs in a pre-configured cloud environment opened from a browser, with no software to install. Participants choose a visual build track or a coding-agent track, and each lab is verified against fixed acceptance criteria. Participants leave with a working triage agent, a service prompt library, an evidence pack of traces and metrics, governance starting points for AI in front of customers, and a ninety-day plan for their function. The programme finishes on where the field is heading: agentic AI, where a support agent is given a goal and plans and executes the work under human supervision, with context engineering, agentic retrieval-augmented generation, the Model Context Protocol and voice agents.

Who should attend

This programme is designed for the leaders and managers of customer service, support and contact-centre teams across all industries:

  • Heads of Customer Service and Customer Experience Directors
  • Contact Centre Managers and Call Centre Managers
  • Customer Support Managers and Support Team Leads
  • Customer Success Managers and Account Support Leads
  • Service Desk and Helpdesk Managers
  • Complaints and Escalations Managers
  • Quality Assurance and Workforce Management Leads
  • Knowledge Managers and Service Operations Analysts
  • Digital Channel, Chatbot and CX Product Owners

Learning outcomes

By the end of this programme, participants will be able to:

  • Explain how generative AI works and where its output needs verification
  • Apply the C.L.E.A.R. framework and K-shot prompting to customer service work
  • Map AI across the service workflow, from triage to follow-up
  • Draft and review customer replies and complaint responses with AI
  • Ground a knowledge assistant in policies, FAQs and resolved tickets
  • Design a customer-facing chatbot with fallback and handover to a person
  • Extract sentiment and themes from ticket history
  • Design a support agent's goal, inputs, outputs and guardrails
  • Build a working ticket triage and first-reply agent
  • Evaluate an agent's traces, metrics, costs and failure modes
  • Take a go or no-go decision on an agent proposed for customers
  • Apply governance controls to customer-facing AI
  • Build a ninety-day AI plan for their service function
Programme outline

Twelve modules. Two days. One working support agent.

Each module follows the TEA™ framework, twenty percent theory, twenty percent examples, sixty percent hands-on application on your own service scenarios. Day one applies AI across the service desk; day two builds, evaluates and governs a support agent.

Day 1: Generative AI Across the Service Desk

01
Day 1 · Module

Generative AI Essentials for Customer Service

  • What is generative AI?
  • Large language models and chat assistants (ChatGPT, Claude, Microsoft Copilot, Gemini)
  • Tokens, context windows and memory
  • Chat models and reasoning models
  • Hallucination and non-determinism
  • Verification of AI output before it reaches a customer
  • Customer data, confidentiality and permitted tool use
Live walkthrough the calculator test
02
Day 1 · Module

Prompting for Service Work

  • The C.L.E.A.R. prompt engineering framework
  • K-shot prompting (zero-shot, one-shot, few-shot, multi-shot)
  • Chain-of-thought prompts
  • Persona, tone and house-style instructions
  • Structured outputs (reply drafts, ticket summaries, tables)
  • Common prompting errors in service replies
  • The service prompt library and reusable agent skills
Hands-on exercise applying C.L.E.A.R. to participants' own customer enquiries
03
Day 1 · Module

AI Across the Customer Service Workflow

  • The service workflow (contact, triage, resolution, escalation, follow-up, learning)
  • Ticket triage and routing
  • Response drafting and the review before sending
  • Escalation summaries and hand-over notes
  • Multilingual customer correspondence
  • Knowledge articles from resolved tickets
  • Themes across ticket history
  • Deterministic and probabilistic work in the service desk
  • Where the human stays in each workflow
Gamified task sorting service desk tasks into scripts, agents and human-only work
04
Day 1 · Module

Response Drafting and Complaint Handling

  • The first reply from policy and past answers
  • Complaint handling guidelines and tone under pressure
  • Apologies, remedies and permitted commitments
  • Policy sources for replies (response-time commitments, service credits, refunds)
  • Consistency across channels (email, chat, web form, social)
  • Review and edit before sending
Hands-on exercise the customer email, drafting and refining a service reply against complaint handling guidelines
05
Day 1 · Module

Knowledge Assistants and Grounded Answers

  • Retrieval-augmented generation (RAG)
  • Grounding answers in policies, FAQs and resolved tickets
  • Source citation and abstention when the answer is not in the source
  • Custom assistants on company documents
  • Chatbot design:
    • Intents, flows and fallback
    • Handover to a person
    • Disclosure and tone
  • Knowledge base maintenance and article ownership
Hands-on exercise grounding a custom assistant in a support policy document and testing it for abstention
06
Day 1 · Module

Sentiment, Themes and Service Analytics

  • Sentiment analysis on tickets, chats and reviews
  • Theme and root-cause extraction across ticket history
  • Voice-of-the-customer summaries and the weekly themes report
  • Quality review of replies at scale
  • Service metrics (first response time, resolution time, first contact resolution, CSAT, NPS)
  • Reporting service performance to a management audience
Individual and pair work extracting sentiment and themes from a sample ticket export

Day 2: Building, Evaluating and Governing a Support Agent

07
Day 2 · Module

AI Agents for Customer Service

  • What is an AI agent?
  • Agents, assistants and workflow automation
  • The agent loop (goal, observe, reason, act, evaluate, repeat)
  • Levels of autonomy and human-in-the-loop
  • The build spectrum (no-code, low-code, full-code)
  • Visual agent builders and coding agents (n8n, LangFlow, Claude Code)
  • AI features inside contact-centre platforms (Zendesk, Salesforce Service Cloud, Freshdesk, Intercom)
  • The agent design brief (goal, inputs, tools, outputs, guardrails, success criteria)
Live demonstration an agent completing a multi-step support task end to end
08
Day 2 · Module

Building the Triage and First-Reply Agent

  • The lab environment (GitHub Codespaces, browser-based, pre-configured)
  • The support inbox case (tickets, categories, priority rules, routing teams)
  • Ticket categories (Billing, Technical, Account, Feature Request, Complaint)
  • Routing teams (Finance Desk, Engineering Support, Customer Success, Product, Escalations)
  • Priority rules (P1 to P3, by customer plan and issue type)
  • First-reply drafting for priority-one tickets
  • The two build tracks (visual canvas, coding agent)
  • Planted edge cases (ambiguous, out-of-policy, missing information)
  • Acceptance criteria and automated checking of results
Lab building the ticket triage and first-reply agent end to end
09
Day 2 · Module

Monitoring and Evaluating the Agent

  • Observability for AI agents (LangSmith)
  • Traces and their contents (context, prompt, model response, tokens, latency, cost)
  • The metrics baseline
  • Failure-mode probes and abstention on missing information
  • Optimisation levers (instructions, context size, model choice)
  • Measured before-and-after comparison
  • The Agent Health Check measures:
    • Groundedness and source citation
    • Escalation and abstention behaviour
    • Hallucination and omission rates
    • Sensitive-data leakage
    • Cost, latency and reliability
Lab instrumenting the triage agent, recording the baseline and running a failure-mode probe
10
Day 2 · Module

The AI Review Board and Service Governance

  • The review board case (a support agent's pilot output: tickets, policies, replies, run log)
  • Promises to customers against written policy
  • Security-related tickets against the security policy
  • Cost and reliability review of the run log
  • The go with conditions or no-go memo
  • Governance for customer-facing AI:
    • Human approval points and permitted agent actions
    • Customer data, privacy and retention
    • Disclosure of AI use to customers
    • Acceptable-use policy and the agent register
  • The regulatory and standards direction (EU AI Act, ISO/IEC 42001, NIST AI Risk Management Framework)
Case study analysis judging a deployed support agent's work and issuing a go or no-go memo
11
Day 2 · Module

Emerging Methods

  • Context engineering
  • Agentic retrieval-augmented generation (RAG)
  • Model Context Protocol (MCP)
  • Voice agents and real-time speech
  • Computer use and browser agents
  • Multi-agent patterns in the contact centre
Controlled group discussion which emerging methods belong in participants' service desks within twelve months
12
Day 2 · Module

Capstone and Personal Action Plans

  • The service AI use-case portfolio
  • Value, feasibility and risk assessment
  • The pilot definition for the triage agent (scope, owner, success criteria, review points)
  • Team readiness and capability building
  • The ninety-day plan (next steps, priority use case, timeline, owners)
Personal action planning each participant's ninety-day AI plan for their service function
Day 2 · Course closing

Course closing

  • Course debrief and key takeaways
  • Final Q&A
  • Official closing ceremony and award of certificates
  • Group photo
Our training methodology

Three frameworks behind every programme.

Whatever programme you choose, it follows three proven instructional frameworks, from design to delivery to evaluation.

Delivery

TEA™ Delivery Framework

  • T Theory, core concepts & frameworks
  • E Examples, live demos & case studies
  • A Application, hands-on exercises & practice

Our signature approach structures every module into three phases for maximum retention and practical application.

Design

ADDIE Instructional Design

  • A Analyse, assess learner needs
  • D Design, structure the learning path
  • D Develop, build materials & activities
  • I Implement, deliver the programme
  • E Evaluate, measure outcomes & refine

The industry-standard model for course design and learning alignment.

Evaluation

Kirkpatrick Model

  • L1 Reaction, participant satisfaction
  • L2 Learning, knowledge & skills gained
  • L3 Behaviour, on-the-job application
  • L4 Results, business impact & ROI

The globally recognised four-level framework for measuring training effectiveness, from initial reaction to long-lasting post-training effects.

Upcoming sessions

Choose your date.

Private cohorts for customer service and support teams, live online or on site. Schedule a TNA call and receive a custom proposal within 48 hours.

Date
Format
Duration
Price
Private cohorts run on request.
Live virtual or On-site
1/2 to 5 days
Custom
What every participant gets

You keep everything.

No vendor lock-in, no paywalled materials. Every delegate leaves with the working files, the templates, and the post-programme support to actually use them.

Personal prompt library

8 – 12 tested prompts built during the session on your actual work scenarios.

Templates & dashboards

Dashboards, workflow templates, and evaluation frameworks ready to use.

Learning materials

Full soft copy of all session materials, handouts, and reference guides.

STOA certificate

Certificate of Completion plus a guided pathway to the Google AI certificate.

30-day support

Post-programme Q&A availability and implementation guidance.

Lifetime updates

Access to updated prompt libraries and resources as AI tools evolve.

Delivery options

Choose how your team learns.

In-Person

Instructor-led on-site

Delivered at your offices or venue. Maximum engagement for teams of 8 – 25 participants. All exercises use your team's actual work scenarios.

Live Online

Virtual instructor-led

Real-time delivery via Zoom, Meet, or Teams. Interactive breakout rooms, same content depth as in-person. Ideal for distributed teams.

Hybrid

Blended approach

Combine in-person and virtual sessions. Flexible scheduling across time zones. Pre-work materials and follow-up resources included.

Training catalogue

104 courses in nine families, overview level, methodology up front.

Download the catalogue
Ready to build

Give your service team a working agent and the judgement to run it.

Book a private programme for your customer service team. Custom proposal within 48 hours. Delivery live online or on site.