LinkedIn Intelligence | She Leads AI

LinkedIn has years of your conversations and the words your audience uses for its own work. This skill reads that archive and hands back the people who already told you they want in.

What It Is

You request your data archive from LinkedIn and hand the unzipped folder to the skill. It reads your message history first, because that is where people said things like “I want to come to your conference.” Then it drafts your Priority 1 list, the twenty to fifty people worth strategic attention this quarter, and your engagement list, the accounts whose audiences overlap yours. Last it counts the phrases your connections use in their own headlines and shows you which of those words never appear in your profile.

What Comes Back

Warm leads — every person who expressed interest in an event, a program, or working with you, with her exact words and the date. You reply to what she wrote.
Priority 1 — twenty to fifty people to engage strategically, each with one line on why she is on the list.
The engagement list — the accounts to comment under, ranked by how much their audience looks like yours.
Relationships cooling — conversations that were in motion and went quiet, sorted by how much was on the table.
The language your network uses — the phrases your audience reaches for, counted and grouped, next to a column showing whether each one is in your profile today.

Why the Messages Come First

When Anne ran her own archive through Claude, the message file held people saying outright that they wanted to attend CREATE. Those conversations had been invisible for months. The connections file tells you who is in your network. The messages tell you who already asked.

Then What

The skill folder includes the playbook that goes with the lists. Put the engagement list in a bookmark folder and go through it on a recurring calendar block. Be early in their comments, with substance. Choose one demand generation channel and commit to it for ninety days. The playbook covers strategic posting, live events, audience borrowing, quiz and webinar funnels, and the portfolio page people find when they search for you.

Make It Yours

The Tier 1 questions ask who is a fit for your event and your flagship program. Swap in yours. Add your own do-not-contact list to the context files. The rest runs as written.

Where This One Comes From

This skill is the LinkedIn Intelligence + Demand Generation Playbook from the Certified AI Educator Program at She Leads AI, Module 4, taught February 19, 2026, rebuilt as a skill so the analysis runs the same way every time.

Make it yours. Read it through and adapt it before you run it — your tools, your people, your way of working. It is built to be changed.

--- name: linkedin-intelligence version: 1.0.0 description: > Use when the user says "analyze my LinkedIn data," "mine my LinkedIn export," "who in my network wants to work with me," "find my warm leads on LinkedIn," "who should I be engaging with on LinkedIn," "what words do my connections use," or "build my engagement list." Also fires on "I downloaded my LinkedIn data, now what," "who already told me they want in," or "sponsorship prospects in my network." Turns a LinkedIn data export into your warm leads and your engagement list, in the words your audience uses for itself. --- # LinkedIn Intelligence v1 Reads a LinkedIn data export and returns the people who already said they want what you offer and the words your audience uses to describe its own work. ## Context Required Read these files before running - The Three Tiers of Questions, below in this file, in priority order - Paste-Ready Prompts, below in this file, one per pass - Where the Lists Go, below in this file, once the report exists - Your own do-not-contact list, if you keep one - Your brand and voice rules, if your team keeps them in a file Data source is the LinkedIn archive you requested at Settings → Data Privacy → Get a copy of your data, unzipped. Ask for the larger archive. LinkedIn emails you when it is ready. For enriched fields (About text, follower counts, tags), use an export from a tool such as LeadDelta rather than scraping. ## Steps 1. **Inventory the export** (TIGHT). List every CSV in the folder with its row count. The ones this skill uses are `Connections.csv`, `messages.csv`, `Invitations.csv`, `Profile.csv`, `Skills.csv`, `Comments.csv`, `Shares.csv`, and `Reactions.csv`. Say which are present and which are missing before analysis starts. 2. **Clean the connections file** (TIGHT). `Connections.csv` opens with a short note above the header row. Skip to the line that starts with `First Name`. Columns are First Name, Last Name, URL, Email Address, Company, Position, Connected On. Many rows have a blank Email Address. That is LinkedIn's default and it is fine. 3. **Mine the message history first** (TIGHT). This is where the buying signals live. Read every row of `messages.csv` where FROM is someone other than the user. Flag any message that expresses interest in an event, a program, working together, sponsoring, speaking, or learning more. For each flag, record the person, the date, and her exact words. Never paraphrase the signal. 4. **Run the Tier 1 questions** from the tiers below against connections plus messages. Tier 1 finds prospects, advocates, cooling relationships, and dormant conversations worth reopening. 5. **Draft the Priority 1 list** (LOOSE on selection, TIGHT on size). Twenty to fifty people. Selection weighs the message signal first, then title and budget authority, then fit with the audience the user serves. Each row gets one line on why she is on the list. 6. **Draft the engagement list**. Twenty to fifty accounts whose audiences overlap with the user's. These are people to comment under, not people to sell to. Pull candidates from the connections file and from names that recur in Comments.csv and Reactions.csv. 7. **Run the language pass**. The export carries Position and Company, so headline language is what is available. If an enriched export with About text is present, use it. Count the most frequent phrases and group them into themes. Compare against the words in `Profile.csv` and the user's recent `Shares.csv` and list the gaps. 8. **Run Tier 2 and Tier 3** only if the user asks or if time allows. Tier 1 is the strategic value. The rest is context. 9. **Apply the do-not-contact list and the names rule**. Remove anyone on the list. Strip the Email Address column from anything that leaves the session. Names on the lists are for the user's follow-up, never for publication. 10. **Run your brand and voice check** on every sentence of the report before showing it. 11. **Present the report** in the format below, then ask the first-run check-in questions. ## Output ``` # LinkedIn Intelligence Report — [First Last] — [MM.DD.YY] Export dated [date]. [N] connections, [N] conversations, [N] invitations. ## Warm Leads — they already told you | Person | Date | Her words | Signal | Suggested next move | |---|---|---|---|---| ## Priority 1 — twenty to fifty people to engage strategically | Person | Title, Company | Why she is here | Last contact | |---|---|---|---| ## Engagement List — comment under these accounts | Person | Why her audience overlaps yours | |---|---| ## Relationships Cooling | Person | Last exchange | What was in motion | |---|---|---| ## The Language Your Network Uses | Phrase | Count | Theme | In your profile today | |---|---|---|---| ## Network Composition (Tier 3, on request) Top industries, top companies, connection growth by quarter. ## Patterns Worth Acting On Two to four sentences. What people keep asking for, and what that suggests for the next thirty days of content. ``` Every table row is one line. No row without a source cell the user can check back in the export. ## Works With **Chains upstream from** - A context or profile file about the user's offers (OPTIONAL) — when present the skill knows what she sells and can score fit. Without it, the skill asks two questions about what she sells and who buys it before Step 5. **Chains downstream to** - A sponsorship or sales pipeline skill — the sponsorship-prospect rows from Tier 1. - A content or hook-writing skill — the language table, so posts use the audience's words. - A relationship follow-up skill — the Relationships Cooling table, for the cadence. - A pitch skill — the Priority 1 list, when the user is pitching herself to the people on it. **Pairs well with** - A devil's-advocate or critique skill — pressure-test the Priority 1 list before the outreach starts. - A chief-of-staff agent — hand it the follow-up list so the touches get scheduled. **Standalone mode** The full report above from the export alone. The lists are usable the same day. ## Gotchas I know you'll want to start with the connections file because it is the biggest — don't. Start with messages.csv. The connections file tells you who is there. The messages tell you who already asked. I know you'll want to read `Connections.csv` with a default CSV loader — don't. It has a note above the header and the loader will treat the note as the header row. Skip to the `First Name` line. I know you'll want to call the Position field a bio — don't. The export has headlines, not About text. Say "headline language" unless an enriched export is present. I know you'll want to summarize what a warm lead said — don't. Quote her. The user will reply to her, and the reply has to reference the exact thing she wrote. I know you'll want to rank the engagement list by follower count — don't. Rank by audience overlap. A person with a small audience of exactly the right women outranks a big general account. I know you'll want to put every good name on the Priority 1 list — don't. Fifty is the ceiling. The list is for weekly attention and nobody gives weekly attention to two hundred people. I know you'll want to hand back a spreadsheet with the Email Address column intact — don't. Strip it before anything leaves the session. The user has those addresses in the export if she needs them. I know you'll want to skip the do-not-contact check because the list is only for the user — don't. Every list leaves the session eventually. Add to this list as failures surface. ## Constraints - Warm-lead signals are quoted verbatim and dated, with the person named. Never inferred from a like or a profile view. - Priority 1 and the engagement list are twenty to fifty rows each. Never more. - Nothing from `messages.csv` is published, posted, or pasted into a shared workspace. The archive stays local. - Names from the report never appear in public content. - No follower counts, connection counts, or audience sizes in anything public-facing. - The skill reads the export. It never scrapes LinkedIn and never sends messages or connection requests. ## Check-Ins After the first run, ask - Are you editing the report? Which table? - Is anyone on Priority 1 who should not be, or missing who should be? - Did the warm-lead quotes match what you remember of those conversations? Before wrapping up, ask - Any failure pattern to add to the gotchas while it is fresh? - What else can I take off your plate? ## The Three Tiers of Questions Run them in order. Tier 1 is where the strategic value is. Tier 2 and Tier 3 add context once the lists exist. ### Tier 1 — Prospects, warm leads, and relationships - Who are the top sponsorship prospects in the network? - Who has expressed interest in attending your event? - Who is a strong fit for your flagship program? - Who are the most likely advocates or referrers? - Which connections show the highest engagement with your posts? - Which important relationships show declining engagement? - Which dormant conversations should be reopened? - Who are the most influential people in the network? ### Tier 2 — Patterns, demand signals, audience behavior - What themes of professional interest appear most often? - What professional archetypes show up across the network? - What share of the network represents active professional relationships? - What is the balance of inbound and outbound connection activity? - What is the balance of reciprocity and value exchange? - Is there a relationship between how often you post and how fast the network grows? ### Tier 3 — Composition and growth - What is the demographic composition of the network? - Which companies are most represented? - Which industries are most represented? - What are the historical growth trends? - Which skills, keywords, and phrases are associated with the professional brand? ### Where each answer comes from | Question group | Files | |---|---| | Interest, warm leads, advocates | messages.csv, Invitations.csv | | Engagement, cooling, influence | Comments.csv, Reactions.csv, messages.csv | | Themes, archetypes, language | Connections.csv (Position), Profile.csv, an enriched export if present | | Reciprocity, inbound vs outbound | Invitations.csv (Direction column) | | Composition and growth | Connections.csv (Company, Position, Connected On) | ## Paste-Ready Prompts Each prompt runs against one file or one pass. Use them in order. Paste into whichever AI you use with the file attached. ### Pass 1 — Warm leads from message history Attach `messages.csv`. ``` Read every message in this file that was sent to me, not by me. Find anyone who expressed interest in attending an event, working with me, sponsoring something, speaking somewhere, or learning more about what I do. For each person give me a table with her name, the date, her exact words in quotes, the kind of interest, and one suggested next move. Quote her. Do not paraphrase. Then tell me what people keep asking about, in two to four sentences. ``` ### Pass 2 — Priority 1 list from connections Attach `Connections.csv` and the table from Pass 1. Tell the AI who you serve and what you sell in one sentence each. ``` This file has a note above the header row. Skip to the line that begins with First Name. I serve [who] and I sell [what]. From these connections, pick twenty to fifty people I should engage with strategically. Weigh people from the warm-lead table first, then titles that signal budget authority, then fit with who I serve. Give me a table with name, title and company, why she is on the list in one line, and the date we connected. Leave the email column out. ``` ### Pass 3 — Engagement list Attach `Connections.csv`, `Comments.csv`, and `Reactions.csv`. ``` Find twenty to fifty accounts whose audiences overlap with mine. These are people whose posts I should comment under, not people I am selling to. Weigh the people I already interact with in the comments and reactions files, then connections whose headline describes the same audience I serve. Rank by audience overlap, not by follower count. One line each on why her audience overlaps mine. ``` ### Pass 4 — The language your network uses Attach `Connections.csv` and `Profile.csv`. Attach an enriched export with About text if you have one. ``` Count the most frequent phrases in the Position column, or in the About text if it is present. Group them into themes. Then compare against the words in my own profile and tell me which phrases my network uses that I never use. Give me a table with phrase, count, theme, and whether it appears in my profile today. ``` ### Pass 5 — Relationships cooling Attach `messages.csv`. ``` Find conversations that were active and then went quiet. For each, give me the person, the date of the last exchange, and what was in motion when it stopped. Sort by how much was in motion, most first. ``` ## Where the Lists Go — Engagement and Demand Generation The report produces lists. This is what to do with them. ### The engagement list, in practice The algorithm rewards engagement. The point is the relationships, and for your top-tier targets you go around the algorithm. 1. Put the twenty to fifty accounts from the report in a browser bookmark folder called LinkedIn Engagement List. 2. Block a recurring calendar slot to go through the folder. 3. When one of them posts, be early in the comments. Comment with substance. "Great post" is not a comment. 4. The goal is to become a recognized name in their comment threads. Their audience becomes aware of you. On Premium. If you have it, your profile views are hidden from the people you visit. If you do not, it changes little. Most people are doing the same research you are. ### Demand generation Demand generation is building interest before someone knows she needs you. Momentum compounds. The flywheel is hard to start, and once it moves, it moves. #### Channel 1 — Strategic social media Posting is table stakes. The strategic layer is knowing who you post for and making sure they see it. - Post on LinkedIn a few times a week. - Pick topics from the language table in the report. Write about what your audience is already thinking about, in the words they use. - Tag people when the content connects to them. Never in bulk. - Your engagement list is how the content reaches the right eyes without paid reach. #### Channel 2 — Live events, in person and virtual Live events accelerate trust faster than any other format. - Go live on LinkedIn with a collaborator in your space. LinkedIn notifies both audiences. - Host or co-host virtual events that give your community a reason to gather. - Guest at other people's events. You borrow their audience and they borrow your expertise. - In person is back. When the opportunity comes, say yes. #### Channel 3 — Audience borrowing Trading audiences grows your reach without the overhead of doing everything alone. - Find podcasts in your adjacent space and pitch yourself as a guest. - Guest-author in newsletters that reach your audience. - Co-create workshops with complementary practitioners. You split the work and both reach new people. - Whenever you guest anywhere, have a landing page you control ready for the new contacts. #### Channel 4 — Funnels Two flows worth modeling. Start with one. **Quiz funnel.** A short quiz tied to a pain point your audience has. The answers tell you where each person is in her journey. Platforms like Interact or Typeform do this. Drive traffic from LinkedIn posts, email, or a small ad budget. A completed quiz is a new contact, followed by a targeted offer. **Webinar funnel.** A free training on one specific, high-demand topic. Registrations come from LinkedIn, email, or low-cost ads. End with one clear offer, such as a consultation or a program. Record it and repurpose the replay. #### Channel 5 — Your portfolio presence Before someone hires you or joins your community, she searches for you. Make sure what she finds is worth landing on. - A simple landing page that answers who you are and where to go next. - Your speaker bio, your programs, one way to get in touch, and one thing to download. - Gamma produces a polished page fast. A coding tool like Windsurf works if you want to build it yourself. - Get something live this week. Add analytics or a link tracker so you know if anyone visits. ### Implementation sequence Each step builds on the last. 1. This week — request your LinkedIn data archive. 2. This week — run your message history through the skill. Find your warm leads. 3. This week — build the engagement list and bookmark it. 4. Next week — start the recurring engagement block. 5. Next week — get one portfolio page live. 6. Thirty days — choose one demand generation channel and commit to it for ninety days. 7. Sixty days — decide on the quiz funnel or the webinar and build it. The goal right now is visible. Every connection you deepen and every comment you leave is the flywheel turning. ### Questions to take further - Who in your connections already gave you a buying signal you have not followed up on? - If the language pass revealed a word you never use, what would happen if you built your next thirty days of content around it? - Which five people on LinkedIn have audiences that look like your ideal client, and what would it take to be in their comment sections this week? - If the algorithm reset tomorrow, which relationships would still send people your way? - Which one channel, if you committed to it for ninety days, would most likely produce a measurable result? ## Changelog | Date | Version | What Changed | |------|---------|-------------| | 09.02.26 | 1.0.0 | Initial skill. Built from the LinkedIn Intelligence + Demand Generation Playbook, Certified AI Educator Program at She Leads AI, Module 4, February 19, 2026 session. | --- © She Leads AI · hello@sheleadsai.ai · Sharing encouraged.