This lesson is being recorded

Every video of this course is being uploaded to my YouTube channel. Subscribe so you do not miss it.

Bruno AshfordBruno Ashford @brunoashfordSubscribe

Module 8 · Lesson 8.6 · 12 min

Customer Service Playbook

Draft replies with the right tone, build a living FAQ, and escalate anything sensitive to a human.

Module 8
Revenue Playbooks

Sales, marketing and customer service that run on your AI department

Bruno Ashford
Bruno Ashford · Module 801 / 13
Why start here
Put your AI department where the money is first. Every hour saved here is an hour closer to a customer.
Bruno Ashford
Bruno Ashford · Module 802 / 13
The six playbooks

From stranger to happy customer

  1. 1Lead researchKnow them before you call
  2. 2Follow-upNo lead goes quiet
  3. 3ProposalsSame-day, tailored
  4. 4ContentOne idea, a week of posts
  5. 5CampaignsCost per customer, not clicks
  6. 6ServiceRight tone, right escalation
Bruno Ashford · Module 803 / 13
8.1 Lead research

Research stops being the skipped step

The researcher cites a source for every fact.

  • Sharp ideal customer profile first
  • Parallel batches of five
  • Contact fits of 4 and 5
  • Public business information only
Bruno Ashford · Module 804 / 13
8.2 Follow-up
Leads rarely say no. They go quiet when you do.
Bruno Ashford
Bruno Ashford · Module 805 / 13
8.2 The morning routine

15 minutes that close the leak

CRM in, drafts out, you send.

  • Overdue list, read-only
  • A specific reason to reply
  • Under 80 words each
  • Log it, set the next date
Bruno Ashford · Module 806 / 13
8.3 Proposals

Transcript to proposal

  1. 1CaptureCall notes or transcript
  2. 2DraftSkill plus pricing rules
  3. 3ReviewReviewer catches blockers
  4. 4DecideYou resolve every [CONFIRM]
Bruno Ashford · Module 807 / 13
8.4 Content engine

One real idea, four formats

Consistency beats inspiration.

  • Ideas file from real customers
  • Voice written in CLAUDE.md
  • Post, caption, email, script
  • Human approves before publishing
Bruno Ashford · Module 808 / 13
8.5 Campaigns (illustrative numbers)

Ask the business question

CampaignCost per leadCost per customerVerdict
Branded search$18$42Scale
Meta prospecting$9$190Fix or pause
Retargeting$14$65Keep

Illustrative example. Cheap leads are not cheap customers.

Bruno Ashford · Module 809 / 13
8.6 Customer service

Always escalate these to a human

Write these rules before drafting a single reply.

  • Refunds and money
  • Legal or safety issues
  • Health questions
  • Strong emotion, public complaints
Bruno Ashford · Module 810 / 13
Measure it (illustrative targets)

What owners typically track

5 minresearch per lead, down from 20
0leads without follow-up after 7 days
Same dayproposal turnaround
Bruno Ashford · Module 811 / 13
Project

One playbook, end to end, measured

Lead, research, contact, follow-up, proposal.

  • Log hours before you change anything
  • Run it for two weeks
  • Log hours after
  • Turn hours into money
Bruno Ashford · Module 812 / 13
Recap
AI drafts, researches and remembers. You decide, relate and close. That is the split that grows revenue.
Bruno Ashford
Bruno Ashford · Module 813 / 13
01 / 13

Use ← → to move. The marker highlights text and the pencil draws on the slide; navigation locks while you annotate. Undo removes the last mark, Esc exits.

1Write your escalation rules first

Before any drafting, list what always goes to a human: refunds and money, complaints about staff, legal or safety issues, health or medical questions, data or privacy requests, and any message with strong emotion. Put this list in CLAUDE.md or a customer service skill.

2Build the FAQ from real messages

Export 50 to 100 recent customer messages or emails and let Claude find the questions that repeat. This becomes your FAQ and the source for routine answers.

Prompt for Claude
Read the customer messages in support/messages-export.csv. Group them into the 15 most common questions. For each, write an approved answer in our brand voice using only facts from CLAUDE.md and our policies in support/policies.md. Mark any answer where the policy is unclear with [OWNER TO DECIDE]. Save it as support/faq.md.
3Triage and draft the inbox

Each day, ask Claude to sort new messages into routine, needs a human, and urgent. Routine ones get a draft reply based on the FAQ. Everything else gets a short summary for the right person.

Prompt for Claude
Read today's new customer messages. Do not send anything. Sort them into ROUTINE, HUMAN and URGENT using the escalation rules in CLAUDE.md. For ROUTINE, create a draft reply using support/faq.md. For HUMAN and URGENT, give me a two-line summary, the customer's emotional state, and who on the team should handle it.
4Reply to reviews with care

For public reviews, ask for a reply that thanks, acknowledges the specific issue, avoids arguing and sharing private details, and moves the conversation offline. A human reads every public reply before it is posted.

5Keep the FAQ alive

Once a month, ask Claude to compare new messages with the FAQ and propose additions and updates. Questions that keep coming up may point to a problem in your product, website or process, which is worth more than any reply.

One angry review can cost you more customers than a month of ads brings in. And the reply to that review is usually written by a tired owner at 11pm. You know how that goes.

The customer service playbook gives every message a calm, on-brand first draft in minutes, answers routine questions from a living FAQ, and, most importantly, knows what it must never handle alone: refunds, legal threats, safety issues, health questions and anyone who is upset. Those go straight to a human.

Key points

  • Write the escalation rules before you draft a single reply.
  • Build the FAQ from real messages, not guesses.
  • Money, legal, safety, health and emotion always go to a human.
  • Repeated questions reveal problems worth fixing at the source.

Avoid

  • Letting the agent reply to angry customers without a human reading it.
  • Answering from general knowledge instead of your actual policies.
  • Sharing a customer's private details in a public review reply.
Real-world example: a property management company

A property management company with 300 units received around 60 tenant messages a day. The playbook sorts them each morning: about two thirds are routine (bin days, parking, how to report a repair) and get drafts from the FAQ. Maintenance emergencies and anything mentioning damp, gas or safety are marked urgent and go straight to the on-call manager. The office manager now clears routine messages in under an hour instead of most of the morning.

Bruno Ashford Subscribe