How to build an AI sales playbook for LinkedIn
Turn your manual sales scripts into a structured playbook that AI agents use to write personalized LinkedIn messages and answer common objections.
Get startedShort answer
An AI sales playbook is a set of rules, goals and objection handlers that tells an AI agent how to write your LinkedIn messages. It gives the agent context about your offer, your buyer and your tone so the writing sounds like a person instead of a bot.
Define your core offer and ideal customer
An AI agent can't guess your value. Give it a vague description and it writes generic corporate language that people scroll past. The playbook starts with a factual foundation: what happens when a customer uses your product, who buys it, and what they were doing before they found you.
Skip the adjectives that belong on a brochure. Don't call your software powerful or intuitive. Describe the specific action it performs and the change the buyer sees. The agent needs the gap between where the customer is now and where they want to be, because that gap is what the message is actually about.
If you sell more than one thing, write a separate block for each offer. Mixing two offers in one block produces messages that hedge between them, and hedging reads as uncertainty. Keep the block to four lines: product, pain, value, target.
The product
- Element
- One sentence describing the service without fluff.
- What to provide
- We book 10 to 15 qualified sales meetings a month using LinkedIn.
The pain
- Element
- The specific problem the buyer has this week.
- What to provide
- Founders spend most of the week messaging strangers and getting no replies.
The value
- Element
- The result that shows up first.
- What to provide
- The founder stops prospecting and only spends time on sales calls.
The target
- Element
- Job title, company size and industry of the buyer.
- What to provide
- CEOs of B2B SaaS companies with 10 to 50 employees.
Map your conversation flow
A playbook covers a sequence of goals, one per interaction. Tell the AI what the objective is at each stage so it doesn't try to close a deal in the connection request.
The first message aims at a reply or an acceptance. The second establishes a problem the buyer recognizes from their own week. The last moves the conversation to a calendar. Each stage gets its own length budget: a connection note has under 200 characters, a first DM has two or three sentences, a follow-up has one new piece of information.
Write the follow-up as a separate goal, not a repeat of the first message. A reminder that says the same thing again gets ignored. A follow-up that adds an observation about their hiring, their launch or their pricing page gives the person something to respond to. If you want ready-made opening lines, these connection notes for founders and CEOs show the shape.
- 1
The hook
Give a reason for reaching out that comes from their profile, a recent post or a job listing. One sentence. This proves it isn't a mass blast.
- 2
The bridge
Connect their role to the specific problem you solve. Say why it matters to them this quarter, not to businesses in general.
- 3
The ask
Use a low-friction request. Don't ask for a 30 minute demo. Ask whether they handle a specific problem, or whether a short call is worth it.
- 4
The follow-up
Add a tip, a number from their own site or a relevant observation. Never send a message that only asks whether they saw the last one.
Build your objection handler library
AI agents fail at the first no. Without a handler they either apologize and stop or push harder, and both end the conversation. Write responses for the four or five objections you actually hear: no budget, already have a vendor, not the right person, send me info, and not now.
A handler acknowledges the concern, then asks a question that qualifies or reframes. Don't argue with the objection. If someone says the timing is wrong, ask what would need to be true for the timing to be right. If someone says they already have a tool, ask what they wish it did better.
Keep each handler to two sentences. Long rebuttals read as defensiveness, and a defensive message is the one people screenshot. The handler should sound like something you'd say to a colleague who just told you they're swamped.
Generic AI response
- I understand you're busy. Please let me know when you have time for a demo.
- We are actually more affordable than our competitors if you look at our pricing page.
Playbook-driven response
- That makes sense. Most founders I talk to are underwater right now. Would it be easier to check back after your Q3 planning?
- Understood. Several clients used that tool before switching because they needed better lead scoring. Are you getting the quality of leads you want from it?
Set the brand voice and constraints
To keep the AI from sounding like a bot, give it a list of forbidden behaviors. Most models default to a formal, helpful tone that feels wrong in a LinkedIn DM. Write the constraints as rules the agent can check. Be casual is not checkable. No paragraph longer than two sentences is.
LinkedIn messages read like a note to a peer at a conference. Short, specific, one idea. If a message fills a phone screen, it gets swiped away before the second line. Give the agent a hard sentence limit and a hard question limit, then hold it to both.
One question per message. A message with two questions gets one answer at best, and usually the easier one.
Paste this job
Tone constraints: - Use contractions (I'm, don't, we've). - No paragraph longer than two sentences. - No emoji in the first message. - Never use the words innovative, cutting-edge or world-class. - One question per message, at the end. - No greetings like Dear or Sincerely. - Never mention pricing in the first two messages.
Test the playbook on five real profiles
Before you let an agent run on a list, run the playbook by hand on five people you'd genuinely want as customers. Paste their profile into the agent and ask for the connection note and the first DM. Read both out loud.
You're checking three things: does the opening line prove someone read the profile, does the problem sound like something this person says about their own week, and would you send it under your own name. If any answer is no, the playbook is missing context, not the model. Add the missing line and run the same five profiles again.
Keep the five profiles as a test set. When you change the tone rules or add an objection handler, rerun them. It takes ten minutes and catches the change that makes every message start with the same sentence.
Example
- You
Saw you're hiring a second SDR. Most heads of sales I talk to hit a wall around the third hire, where the manual research stops scaling. Are you still building the lists yourself?
- Prospect
Partly. It's the research that eats the time, not the sending.
- Suggested reply
That matches what I hear. The research is the part that doesn't shrink when you add a person. Worth a short call to compare how you're splitting it now?
Doing this with Kinlead
Kinlead takes the playbook you just wrote and uses it as the campaign's sales playbook, which is the script the AI writes from. You set the goal, tone and playbook once per campaign, and the agent applies them to every lead in the group. Leads are scored from 0 to 100 against the customer profile you described, and anything below the fit line is dropped, so the playbook only runs on people who match it.
Messages go out from your own LinkedIn account at most one action every five minutes, inside the daily limits and send window you set. Queued messages can be read and edited on the Actions page before they send, and any campaign can be paused. Replies from every campaign land in one inbox sorted by intent, and you get an email when someone answers.
- 1
Add your playbook to the campaign
Paste your offer block, conversation goals and tone constraints into the campaign's playbook field, then pick a tone.
- 2
Read the queue before it sends
Open the Actions page and edit any message that drifts from your voice. Edits apply to that message, not the whole campaign.
- 3
Choose how replies are handled
Hand off at the first reply, let the AI reply until there's real interest, or let it reply until a meeting is booked with your scheduling link.
- 4
Point campaigns at a lead group
Build the group from a described customer or your website, then aim any campaign at it. Groups and campaigns stay separate, so one group can feed several campaigns.
Frequently asked questions
Related
Get started with Kinlead
Prefer markdown? How to build an AI sales playbook for LinkedIn.md