Back to catalog
AI at workBeginner

AI for Sales and Customer Support

Preparing for a deal, first emails, objections, proposals, support replies and escalation — and the promises a model must never make on your behalf

AI for Sales and Customer Support

About this course

A course for the people who talk to one specific customer: sales and account managers, support agents, and owners of small companies who do both jobs at once. It is about the part of the work that happens between you and a named person — reading what is publicly known about a company before the first call, writing an opening email that is not a mail merge, taking an objection apart instead of firing back a script, keeping a deal alive through three months of silence, and answering a support ticket in a way that survives being quoted back at you.

Every lesson carries a prompt you can copy, usually as a pair: a weak request and a stronger one for the same job, with a plain explanation of what the difference buys you. The same amount of attention goes to the opposite question. Generated correspondence fails in specific, recognisable ways: identical letters that recipients learned to skip, warmth that nobody feels, an answer to a question that was not asked, and a promise the company will not keep. The course names each of them and shows what to put in the prompt so they do not reach the customer.

One rule runs through the whole course and does not bend: a model does not promise a customer a date, a discount, a refund or a guarantee that no person has confirmed. A promise from your chat bot is a promise from your company. No programming is required, and nothing here is tied to a particular CRM or to a version of a model.

Course curriculum

  1. Before the conversation starts

    What can be learned about a client before the first contact, how to turn it into hypotheses and questions, why a templated blast stops working, and how to take an objection apart instead of answering it from a script.

    • 1.Reading up on a client without inventing them13 min
    • 2.The first email, and why a template blast fails12 min
    • 3.An objection is a thing to take apart, not to deflect14 min

  2. A deal that runs for months

    Restoring the history of a long thread, naming the next step, reminding without becoming a nuisance, and building a proposal out of the client's own description of their task.

    • 1.The long thread: history, next step, reminder13 min
    • 2.A proposal written out of the client's own words14 min

  3. Support: the request, the tone, the limit

    Answering from a knowledge base instead of from the model's memory, taking an angry message apart before replying, an apology that admits only what happened, and the four kinds of request that never get an automated answer.

    • 1.Answers that lean on the knowledge base12 min
    • 2.The point of the request, the tone, and an honest apology14 min
    • 3.Escalation: requests a model does not handle at all12 min

  4. Bots, the final check, and customer data

    What a site chat bot can and cannot be trusted with, why its promises are the company's promises, the short check before every send, and what must never be pasted into somebody else's service.

    • 1.A chat bot on the site: what it can do and where it breaks13 min
    • 2.The check before you press send11 min
    • 3.Customer data: what does not go into somebody else's service11 min