Build, supervise and improve an AI agent on WhatsApp

Objectives

• Read an agent's conversations over a given period and classify failures by type: wrong answer, missed handover, out of scope, outdated data
• Update the agent's knowledge base (offers, prices, documents, standard answers) and check on a set of test questions that the correction is taken into account before publication
• Adjust the agent's instructions (tone, rules for handover to a human, permitted actions) without widening its written scope
• Track three indicators every week (out-of-hours response rate, handover-to-human rate, qualified leads or resolved tickets) and decide on a corrective action
• Handle an escalation: take over a conversation, close the case, document the incident

Audience

Managers and supervisors in customer service, after-sales, sales and marketing; product managers; anyone appointed as the owner of a custom-built AI agent.

Prerequisites

Use WhatsApp Business or a customer relationship tool daily. No technical skills. Equipment: one laptop per participant with an up-to-date browser; access to a training agent deployed by Yolea on a test WhatsApp number, with its supervision console; for remote attendance, a stable connection of at least 5 Mbit/s, microphone and camera; participants bring their catalogue, their price list and twenty real customer questions.

Programme

• Supervising a conversational agent
• Maintaining a knowledge base
• Controlled configuration of instructions
• Steering by indicators
• Managing handovers and incidents

Format

2 days, 14 hours. In-company, on your own agent or on a training agent; on site, live remote or hybrid; 6 to 12 participants. 2 consecutive days of 7 hours; remote, 4 sessions of 3 h 30 over two weeks. Workshops on a real agent fed with the participants' own documents.

Assessment

Beforehand, a placement questionnaire. During the course, check quizzes and a practical challenge (diagnose and correct ten assigned faulty conversations, with the result measured on the test set). At the end, an end-of-course evaluation and a personal certificate of attendance. At three weeks, a follow-up check-in on the agent's indicators.