How to use sales playbooks for AI outreach
Stop using generic prompts. Learn to write a sales playbook that teaches AI to use your real voice, specific phrases, and concrete nouns on LinkedIn.
Get startedShort answer
Stop using general prompts and start using a sales playbook. A playbook is your actual sales script, containing the phrases and objections you use in real conversations. When you give the AI a script instead of a goal, it mimics your logic and voice instead of guessing what a salesperson sounds like.
Why AI messages fail
Most AI outreach fails because it tries to sound professional. In AI terms, professional means formal, stiff, and vague. When you tell an AI to be professional, it pulls from corporate brochures and press releases. B2B buyers ignore those in their inboxes.
Real buyers respond to directness. They want to know you understand their specific problem and that you aren't a bot sending the same message to a thousand people. If your message uses words you would never say in a real conversation, the reader will spot the AI.
The gap between a bot and a human is in the nouns. Bots use abstract words like solutions, ecosystems, and paradigms. Humans use concrete words like spreadsheets, calendar invites, and leaked leads.
A better sample does more than a better prompt. If you give the AI ten messages you actually sent and they worked, it has a pattern to copy. If you give it a goal like book a demo, it has to guess what a salesperson sounds like, and its guess is usually wrong.
Generic AI
- I noticed your profile and believe our product could improve your workflow.
- We help companies reach their full potential by fixing their lead generation process.
- I would love to hop on a call to discuss how we can build a stronger partnership.
Human sales
- Saw your comment on the post about lead gen. Most people struggle with that same gap.
- We stop lead leakage by automating the first response so you don't lose people to competitors.
- Do you have 10 minutes on Thursday to see if this fits your current setup?
Building your sales playbook
A sales playbook is a collection of your best real-world conversations. Instead of telling the AI to be friendly, give it the exact words you use when you're being friendly. This stops the AI from falling back on a corporate persona.
Review your sent folder. Find the messages that got a response and strip out the prospect's name. What's left is your voice. This includes how you frame a problem and how you move from a compliment to a pitch.
Document how you handle friction. Most AI tools fail during the follow-up because they don't know how to handle a soft no or a request for more information without sounding like a customer service agent.
Your playbook can be messy. It does not need to be a polished document. Copy real messages into a text file, add notes about why they worked, and remove anything that sounds like a brochure. The AI only needs enough examples to see the pattern.
- 1
List your opening hooks
Write 3 to 5 ways you start a conversation based on a trigger. If they started a new job, write the exact phrase you use to congratulate them without sounding like a bot.
- 2
Define the bridge
Write a one-sentence bridge that connects the lead's situation to your solution. Avoid words like optimize or transform. Use plain language like this usually means or the result is.
- 3
Map common objections
List the three most common reasons people say no and write your exact rebuttal. If they have no budget, write the specific way you pivot to a low-friction trial.
- 4
Create a vocabulary list
List words you use and words you forbid. For example, I say chat instead of consultation, or never use the word solution.
Structuring the playbook for the AI
How you organize your script determines how well the AI applies it. If you provide a wall of text, the AI might miss the nuances. Break your playbook into clear categories: tone, value proposition, and call to action.
The tone guide should be a list of constraints. Telling an AI to be casual is too vague. Telling it to use contractions and avoid emojis is a concrete instruction it can follow.
Your value proposition should be a plain statement of fact. Avoid marketing speak. The AI is more effective if it's told to be a direct salesperson rather than a brand ambassador.
Put the strongest example next to each rule. If the rule is keep the first line short, include a first line that is eight words. The AI copies examples more reliably than it follows abstract advice.
Tone guide
- Section
- Adjectives and constraints
- Content
- Direct, skeptical of fluff, no emojis, use contractions.
Value prop
- Section
- The core promise
- Content
- We stop lead leakage by automating the first response.
Call to action
- Section
- How you ask for the meeting
- Content
- Do you have 10 minutes Thursday?
Testing your voice
Once you've fed your playbook into the system, audit the outputs. The first few drafts show if the AI is still leaning on its default training. If you see a word like synergy or paradigm, your playbook isn't restrictive enough.
Use the out loud test. Read the AI's draft out loud. If it sounds like something you'd say to a colleague or a peer at a conference, it's a pass. If it sounds like a script from a 1990s insurance commercial, it's a fail.
Iterate by adding negative constraints. If the AI keeps using the word improve in a stiff way, add a rule that says never use the word improve and replace it with fix or cut.
Example
- Suggested reply
I noticed your recent post and believe our solution could improve your workflow. Would you be open to a brief consultation?
- You
This is too formal. It sounds like a bot. Delete improve and consultation.
- Suggested reply
Saw your comment on the post about lead gen. Most people struggle with that same gap. I have a way to fix it. Do you have 10 minutes Thursday?
If you wouldn't say the sentence out loud to a colleague, delete it from your playbook.
Writing follow-ups that don't sound automated
Follow-ups are where most AI messages fall apart. The first message can hide behind a trigger event. The second message has to sound like a person who remembers the conversation. If your playbook only covers openers, the AI will invent a generic nudge.
Write two follow-ups for every opener. The first follow-up adds a new detail, like a relevant example or a question about their process. The second follow-up gives them an easy exit. Both should be short enough to read on a phone.
Keep the same voice across the sequence. If your opener uses contractions and short sentences, your follow-up should not switch into formal paragraphs. The AI will copy the tone of the examples you provide, so include full follow-up examples in the playbook.
Do not let the AI write a follow-up from scratch. Give it the first message and the follow-up as a pair. That way it can see how the second message refers back to the first without repeating it.
Paste this job
Opener: Saw your comment on the post about [topic]. Most teams I talk to hit [specific problem] right after [trigger]. We fix that by [plain result]. Do you have 10 minutes on [day]? Follow-up 1: One thing I left out: [specific example or detail]. If [problem] is still on your list this quarter, I can send a short breakdown. Follow-up 2: I'll stop here. If [problem] becomes a priority, reply later and I'll check back in a month.
Doing this with Kinlead
You can use this approach with the optional sales playbook field in a Kinlead campaign. AI writes messages and follow-ups from the campaign goal, tone, and your playbook, so it has your script instead of a generic prompt.
You can read and edit queued campaign messages on the Actions page before they send, and any campaign can be paused. Outreach runs from your own LinkedIn account through Unipile, and Kinlead never stores your LinkedIn password. Each account sends at most one action every 5 minutes inside the daily limits and send window you set. You can see the campaign controls at AI LinkedIn outreach.
- 1
Paste your script
Paste your sales script into the sales playbook field in your campaign settings.
- 2
Set goal and tone
Align your campaign goal and tone to match the intent of your playbook.
- 3
Review the queue
Check the Actions page to edit any messages that don't match your voice before they go out.
- 4
Connect your account
Connect through Unipile so sends stay inside your limits and send window.
Frequently asked questions
Related
Get started with Kinlead
Prefer markdown? How to use sales playbooks for AI outreach.md