Go to market / Channel plan

LinkedIn: Profile and Engine

The full build: what the profile says, what an agent can legally touch, what only Tom can do, and the eight weeks that decide whether this channel is worth keeping.

Internal · Prepared for Tom Levi

01 / The honest read

This is a 3 to 6 month compounding play, and starting now is lucky

LinkedIn can produce booked calls for this offer, but not on the timeline anyone wants. Realistically: 0 to 3 free 10-minute calls before 5 October, then 1 to 2 free calls a week from mid-October, converting to 2 to 4 paid ₪250 consultations a month by December, with a first LinkedIn-sourced Viability Assessment somewhere in November or December. Anyone selling a faster number off an empty profile with roughly zero followers is lying.

The binding constraint is not content quality and it is not the algorithm. It is 4 to 5 hours a week of Tom's own attention, sustained for eight consecutive weeks, and the non-negotiable half of that is about 20 minutes a day of manual commenting. If that cannot hold through two live client builds, halve every target in this document rather than pretending.

What will not work, stated flatly

• Posting as a lead source. Nobody goes viral in Hebrew B2B from zero. Content's only job in weeks 1 to 8 is to make a connection request land.
• Cold DMs at volume to Israeli ops and finance managers. They ignore LinkedIn DMs. Reach them through their accountants, their ERP implementers and their own job ads.
• A Company Page as a distribution channel. Organic Page reach is a rounding error. Build it for the logo.
• Buying Sales Navigator in August. The free trial is single use per 365 days. Burning it during "נדבר אחרי החגים" destroys the option for the month it matters.
• Any third-party automation tool. See section 02. This is an account-death risk, not a grey area.
• Reaching the מוסך/יבואן owner-operator at all. He is not here. A plan that pretends otherwise gets sharp for nobody.

What the eight week dead window actually buys

The gap between now and October is the asset. By 5 October the plan produces a finished profile, a connection graph, an artifact library and 18 to 22 published posts. Starting inside the dead window is lucky: it is the cheapest eight weeks of the year to publish your first twenty posts badly.

Judge month one on posts shipped and connection acceptance rate, not revenue. Judge October on DM reply rate against a finished profile.

Contested calls, already decided

02 / What is automatable and what is not

The API buys exactly one useful thing, and it is bigger than expected

A self-serve LinkedIn app in 2026 grants two products and only two: Sign In with LinkedIn using OpenID Connect and Share on LinkedIn (w_member_social). Both are Open Permissions, granted in minutes, no review. That is enough to programmatically publish text, image and PDF carousel posts straight to Tom's personal profile.

The handoff table

03 / Setup, in order

Today, Wednesday 12 August. Tom, about 90 minutes.

Order matters. The Page must exist before the app, and the app must exist before anything is automatable. Experience entries go in after the Page so the logo renders instead of the grey building.

1. Create the Company Page first

   Go to linkedin.com/company/setup/new. Name OctoMonic. Industry IT Services and IT Consulting. Website octomonic.com, size 1-10 employees, location Israel, real founded year. Upload the logo from assets/branding/.

   Tagline, 120 char cap

   Copy

   Automation and custom software for Israeli businesses. Built, deployed and maintained end to end.

   Do not publish it empty. McLead has three reshare posts queued to seed the feed the moment it goes live.

2. Create the developer app

   linkedin.com/developers/apps/new. App name OctoMonic AIOS Publisher. LinkedIn auto-rejects any app name containing a fragment of "LinkedIn" or "Microsoft". Associate LinkedIn Page = OctoMonic. Upload the logo.

3. Verify the app against the Page

   From the Page, go to Admin tools, then Verify, then approve. Tom is super admin approving his own request. The request expires in 30 days if ignored.

4. Products tab, request access to both

   Sign In with LinkedIn using OpenID Connect and Share on LinkedIn. Both are Open Permissions, granted in minutes, no review.

5. Auth tab, then send the credentials to McLead

   Add redirect URL http://localhost:8765/callback, then send Client ID and Primary Client Secret to McLead by DM. McLead writes them to .env as LINKEDIN_CLIENT_ID and LINKEDIN_CLIENT_SECRET and registers them in connections.md. Never paste them into a post, a commit, or a session log.

6. Claim the custom URL

   Priority order: /in/tomlevi, then /in/tomlevi-octomonic, then /in/tom-levi-octomonic. Do not keyword-stuff. /in/tom-levi-ai-automation-israel reads spammy in an email signature and pins him to one category forever.

7. Photo, shot this morning on a phone

   Head and shoulders only, eyes on the upper third, face filling roughly 60% of frame. LinkedIn renders this at about 48px in comments, so anything below the collarbone is wasted. Plain mid-tone or dark wall, never white, which blows out against LinkedIn's UI and makes the head appear to float. Face a window, 10 to 11am light, never a ceiling light. One solid mid-to-dark colour, no logo, no pattern. Real smile.

   Test: shrink it to 48px. If he is not instantly recognisable and approachable, reshoot. Hard no on an AI-generated headshot. The tells are widely recognised in 2026, and for an AI practitioner that is a self-inflicted credibility kill.

8. Paste the profile fields, in this order

   Headline, then About, then Experience times three, then Education, then Projects, then Skills. All copy is in section 04. Experience goes in after the Page exists so the logo renders.

9. Settings pass, 25 minutes
   • Turn ON Providing services. Categories: Custom Software Development, Web Development, IT Consulting, Business Analytics, Project Management.
   • Turn OFF and never enable Open to work. For a founder selling to businesses, that badge says the company is not working.
   • Public profile visibility to full and public.
   • Profile viewing set to your full name, private mode off. "Tom Levi viewed your profile" is a free, legitimate warm touch before any DM.
   • Leave Make follow primary OFF. He needs Connect as the primary button until roughly 1,000 connections, because DMs require connections.

   Creator Mode no longer exists. Do not go hunting for the dead "Talks about" field, because that top-5 hashtag box was removed along with the toggle.

10. Contact info, three website links with custom labels

    Almost nobody uses the label field, which is exactly why it works.

    • Book a free 10-min call to octomonic.com/book
    • 10 live product demos to octomonic.com/demos
    • Sample spec (full document) to octomonic.com/sample/viability-assessment-sample.html

    Add tom@octomonic.com as the profile email.

11. Add the Hebrew profile version

    Profile, then Add profile in another language, then Hebrew. Name field is exactly תום לוי. Never put Tom Levi | תום לוי or a title in the name field. LinkedIn actively flags name-field extras and this account is a business asset.

Today, McLead, in parallel, about 4 hours

I build the system. I also get the alert when it breaks.

אני בונה את המערכת. אני גם זה שמקבל את ההתראה כשהיא נשברת.

04 / The profile

Paste-ready copy

Every block below has its own copy button. What lands in your clipboard is exactly what goes into LinkedIn, dashes and bullet characters included, because LinkedIn strips markdown and those punctuation marks are literal.

Headline, two variants

The headline's job is not search. Nobody is searching "automation Israel" and landing on a 40-connection profile. Its job is winning the click when Tom's name sits in a comment thread next to twenty people calling themselves AI automation consultants, and only about 50 to 70 characters survive truncation in that context.

Restaurant owner turned software builder | I automate the manual work inside Israeli businesses — and you own the spec, whether or not I build it | Founder, OctoMonic | Hebrew & English

I turn "I have an idea" into a build-ready spec you own | Automation & custom software, designed and maintained by me | Founder, OctoMonic | Free 10-min call | Hebrew & English

Automation & custom software for Israeli businesses | Founder, OctoMonic | I design, build, deploy and maintain the system myself | Free 10-min call before you spend anything | Hebrew & English

Hebrew headline, for the Hebrew profile version

This is the restaurant version. It opens מסעדן שהפך למפתח, which is the direct translation of "Restaurant owner turned software builder", so it carries the same decision. If Variant B wins, this string needs rewriting before it is pasted, and the Hebrew About below needs the same paragraph swap.

מסעדן שהפך למפתח | בונה אוטומציה ותוכנה בהתאמה אישית לעסקים בישראל — והאפיון נשאר שלך, גם אם לא תבנה איתי | מייסד OctoMonic

About, two variants

Roughly 2,050 of the 2,600 available characters. The first 209 characters are engineered as the pre-"see more" hook: a concrete verifiable system fact with no self-description in it. That opening is byte-identical in both variants and must not be touched. The variants differ in exactly one paragraph, the third.

Formatting: LinkedIn strips markdown, so the middot bullets are literal characters. Keep them. No emoji. No hashtag line at the bottom, which does nothing and reads as 2019.

Every night, a system I built checks the price of 1,121 products across 8 competitor websites and messages a human the moment a margin breaks. Ten scheduled jobs, each one alerting on failure instead of failing quietly. I designed it, built it, deployed it, and I'm the one who gets the alert when it breaks.

I'm Tom. I run OctoMonic — automation and custom software for businesses in Israel.

Before this I spent about eight years in food, five of them owning and running my own restaurant. That's where I learned what a thin margin feels like on a Tuesday, and why most "digital transformation" dies quietly: it solves a problem nobody was actually paying for.

So I work backwards from the money. What's being re-typed. What's being missed. What breaks when one person is out for two weeks.

What I do, concretely:

· Turn an idea or a messy process into a build-ready technical spec — scope, architecture, stack, real costs, real risks. You own the document. Build it with me or take it to any other developer. That's the deal, not a sales line.
· Build it — web apps, internal tools, data pipelines, scrapers, n8n workflows, WhatsApp automation, dashboards. Next.js, Node, Python, Postgres.
· Keep it running — monitoring, alerting, and someone to call. Automation usually fails in month four, not week one.

On AI: I don't sell "AI strategy." I run a multi-agent system that operates my own business — daily briefs, monitoring, research, drafting, and a 24/7 agent I talk to on WhatsApp. Running it in production is how I know where these systems actually break and what they should never be trusted with. That's what I bring to your AI question.

Two things I'll say on a first call that most people won't: if the process should be fixed rather than automated, I'll tell you. If the build is bigger than the problem, I'll tell you that too.

Where to start: a free 10-minute call at octomonic.com. No deck, no discovery form. You describe what you're dealing with, I tell you whether there's something worth building and roughly what it costs.

Hebrew or English.
tom@octomonic.com

Every night, a system I built checks the price of 1,121 products across 8 competitor websites and messages a human the moment a margin breaks. Ten scheduled jobs, each one alerting on failure instead of failing quietly. I designed it, built it, deployed it, and I'm the one who gets the alert when it breaks.

I'm Tom. I run OctoMonic — automation and custom software for businesses in Israel.

I've put enough systems into live businesses to know which ones survive. The ones that stick are the ones where somebody could tell me what the problem was costing before we wrote a line of code. That's why most "digital transformation" dies quietly: it solves a problem nobody was actually paying for.

So I work backwards from the money. What's being re-typed. What's being missed. What breaks when one person is out for two weeks.

What I do, concretely:

· Turn an idea or a messy process into a build-ready technical spec — scope, architecture, stack, real costs, real risks. You own the document. Build it with me or take it to any other developer. That's the deal, not a sales line.
· Build it — web apps, internal tools, data pipelines, scrapers, n8n workflows, WhatsApp automation, dashboards. Next.js, Node, Python, Postgres.
· Keep it running — monitoring, alerting, and someone to call. Automation usually fails in month four, not week one.

On AI: I don't sell "AI strategy." I run a multi-agent system that operates my own business — daily briefs, monitoring, research, drafting, and a 24/7 agent I talk to on WhatsApp. Running it in production is how I know where these systems actually break and what they should never be trusted with. That's what I bring to your AI question.

Two things I'll say on a first call that most people won't: if the process should be fixed rather than automated, I'll tell you. If the build is bigger than the problem, I'll tell you that too.

Where to start: a free 10-minute call at octomonic.com. No deck, no discovery form. You describe what you're dealing with, I tell you whether there's something worth building and roughly what it costs.

Hebrew or English.
tom@octomonic.com

Hebrew About, for the Hebrew profile version

Constraint most people miss: the secondary-language profile carries only name, headline, About and job titles or descriptions. It does not carry Featured, Skills, Education detail, recommendations or the Services page. So the Hebrew About has to be self-sufficient and carry its own links inline, because a Hebrew reader has no other path to the demos or the spec.

This is the restaurant version, matching Variant A. Paragraph 3 (לפני זה הייתי כשמונה שנים בעולם המסעדנות) is the one that needs replacing if Variant B wins.

כל לילה מערכת שבניתי בודקת מחירים של 1,121 מוצרים ב-8 אתרים של מתחרים, ושולחת הודעה לבן אדם ברגע שמרווח נשבר. עשר משימות מתוזמנות, כל אחת מתריעה כשהיא נכשלת במקום להיכשל בשקט. אני תכננתי, בניתי, העליתי לאוויר — ואני זה שמקבל את ההתראה כשמשהו נופל.

אני תום. אני מנהל את OctoMonic — אוטומציה ופיתוח תוכנה בהתאמה אישית לעסקים בישראל.

לפני זה הייתי כשמונה שנים בעולם המסעדנות, מתוכן חמש כבעלים ומפעיל של מסעדה משלי. שם למדתי איך מרגיש מרווח דק ביום שלישי, ולמה רוב פרויקטי ה"דיגיטציה" מתים בשקט: הם פותרים בעיה שאף אחד לא באמת שילם עליה.

לכן אני עובד אחורה מהכסף. מה מוקלד פעמיים. מה נופל בין הכיסאות. מה נשבר כשאדם אחד יוצא לשבועיים.

מה אני עושה בפועל:
· לוקח רעיון או תהליך מבולגן והופך אותו למסמך אפיון מוכן לבנייה — סקופ, ארכיטקטורה, סטאק, עלויות אמיתיות, סיכונים אמיתיים. המסמך שלך. אפשר לבנות איתי, ואפשר לקחת אותו למפתח אחר. זה התנאי, לא משפט מכירה.
· בונה — אפליקציות ווב, כלים פנימיים, פייפליינים של דאטה, סקרייפינג, תהליכי n8n, אוטומציה בוואטסאפ, דשבורדים. Next.js, Node, Python, Postgres.
· דואג שזה ימשיך לעבוד — ניטור, התראות, ומישהו להתקשר אליו. אוטומציה נשברת בחודש הרביעי, לא בשבוע הראשון.

לגבי AI: אני לא מוכר "אסטרטגיית AI". אני מפעיל מערכת סוכנים שמריצה את העסק שלי — בריפים יומיים, ניטור, מחקר, טיוטות, וסוכן זמין 24/7 שאני מדבר איתו בוואטסאפ. דווקא בגלל שזה רץ בפרודקשן אני יודע איפה המערכות האלה נשברות ומה אסור להפקיד בידיהן.

שתי אמירות שתשמע ממני בשיחה ראשונה: אם התהליך צריך תיקון ולא אוטומציה — אני אגיד. אם הפיתוח גדול מהבעיה — גם את זה אני אגיד.

איפה מתחילים: שיחה חינם של 10 דקות ב-octomonic.com. בלי מצגת, בלי טופס. אתה מתאר מה קורה אצלך, ואני אומר לך אם יש פה משהו ששווה לבנות וכמה זה עולה בערך.

הדגמות אינטראקטיביות: octomonic.com/demos
דוגמה למסמך אפיון מלא: octomonic.com/sample/viability-assessment-sample.html

tom@octomonic.com

Experience, three entries, and no invented engineering job history

Entry 1 · Founder, OctoMonic

Type: Self-employed. Location: Israel, Hybrid. Start date: the real Osek Murshe registration month. Do not backdate. LinkedIn displays the duration, and a backdate is exactly what gets caught in a reference call.

OctoMonic builds and maintains automation and custom software for businesses in Israel. I'm the one doing the work — design, build, deploy, and the phone call when something breaks.

I don't publish client names. Here's what I can describe:

COMPETITOR PRICE INTELLIGENCE — Israeli retail
· 1,121 products price-checked every night against 8 competitor websites
· 8 separate parsers, each a different site structure and a different anti-bot posture
· 10 scheduled jobs, every one of them alerting a human on failure rather than failing silently
· Postgres, Python, a dedicated VM, and a dashboard the client actually opens
· A second project for the same client is signed and in build

HOW I SELL, IN ORDER
· Free 10-minute call — is there anything here worth building
· 30-minute paid consultation (₪250+VAT) — a real technical answer, not a sales call
· Viability Assessment (₪1,500-₪2,500+VAT) — your idea or your process turned into a build-ready spec: scope, architecture, stack, cost, risks. You own it. Build it with me or take it to another developer.
· Build engagement + retainer — implementation, plus someone who keeps it running

STACK
Next.js, React, Node, Python, Postgres/Neon, Supabase, n8n, Vercel, WhatsApp Business API, LLM orchestration (Claude, Gemini, Groq), ZenRows/Apify for data acquisition.

Ten live interactive demos: octomonic.com/demos
A full sample spec: octomonic.com/sample/viability-assessment-sample.html

Hebrew and English. tom@octomonic.com

Entry 2 · Co-Founder, Monesys

Co-founded with Reey Yadin. A creator monetization platform — payments, subscriptions and tiering for creators. Pre-launch: built, in production, not yet publicly launched.

My side: architecture and full-stack build — Next.js, Supabase, Stripe embedded checkout and customer portal, tiered access.

Entry 3 · Owner / Operator, [restaurant name], about 5 years

The plan marks this entry as not optional and not buried. It is the same decision as the headline and About. If Variant B wins on those, decide explicitly whether this entry stays. Keeping the entry while removing the story from the headline and About is a defensible middle position: the history stays verifiable and dated for the ops-manager reader doing a background check, but it stops being the first thing anyone reads.

Owned and ran a restaurant for about five years: hiring, payroll, suppliers, inventory, POS, and the part where you close the month and find out whether the last thirty days were worth it. Also a silent partner in a second restaurant.

Two things carried over into what I do now. I read a P&L before I read a requirements document. And I know what it costs an owner to change how his staff works — which is why I won't automate a process that should be deleted.

Featured, in this order, built in two waves

Desktop shows about 2 cards before horizontal scroll. Mobile shows 1. Positions 1 and 2 take almost all the clicks.

Why the spec beats the demos for position 1: it is the only asset that shows a stranger exactly what they receive at the rung Tom wants them on, and it serves both buyers at once. The idea-stage founder sees a tangible artifact. The ops manager sees rigour and a document he can take to his boss.

What ₪1,500-₪2,500 actually buys you: the whole document

A real viability assessment, start to finish. Scope, architecture, stack, costs, risks. You own it — build it with me or take it to another developer.

Ten working product demos. Real interfaces, sample data, honestly labelled.

Not screenshots. Ten self-contained interactive demos you can click through right now.

OctoMonic — start with a free 10-minute call

Skills, 22 total, three pinned

Pin these three. They appear on the profile and drive search matching.

Business Process Automation is the ops manager's search term. Full-Stack Development answers the idea-stage founder's "can he actually build my thing." Artificial Intelligence is a broad match in the category he genuinely operates in.

Then add the remaining 19

The last three are on the list deliberately. They are the only skills that back up the restaurant entry and the headline, and they make him findable to the operator-minded buyer who would otherwise file him as another developer. They are therefore inside the section 04 decision. If the restaurant goes, Restaurant Management goes with it, and Operations Management plus P&L Management become unsupported claims with nothing on the profile behind them.

Do not add

Soft and abstract skills dilute the matching signal and are what an empty profile looks like. Prompt Engineering dates badly and reads junior next to "runs a multi-agent system in production." Cap at about 22. Do not list 50.

Endorsements: do not chase them. One exception: after each project closes, ask that contact to endorse the two or three specific skills the project actually used.

"Talks about" topics

One profile, two readers

Do not build two profiles and do not hedge the top card. Both buyers are buying the same first rung, which is a straight technical read before spending real money, so headline, photo and the first About paragraph stay single-minded. The fork happens lower down the page.

Expect 80% or more of LinkedIn-sourced free calls in Q1 to be idea-stage. Ops-manager conversions come almost entirely via DM or referral, with the profile acting as a pass/fail gate. Do not read low ops-manager volume as profile failure. Read it as channel reality.

05 / The content engine

Comments are the channel. Posts are the landing page people hit after clicking his name.

For the first 8 weeks the ratio is 10 comments per 1 post. Everything else in this section is downstream of that.

Five pillars, weighted

Weekly mix at 3 posts: one from pillar 1 or 2, one from pillar 3, one from pillar 4 or 5. At the 2-post floor: one from pillar 3, one from pillar 1 or 2. Pillar 5 sits inside the section 04 decision.

Language rule: roughly 70% Hebrew and 30% English, allocated by pillar, never mechanically alternating, and never bilingual inside one post, because RTL and LTR mixing renders badly and halves the hook. Never the same content in both languages in the same week. If he wants both, space them 10 to 14 days and rewrite from scratch, never translate. The one legitimate bridge: publish in English, then post a Hebrew summary as the first comment, which also gives him a real reason to be first in his own comments.

Format ranking for 2026

Carousel spec: 8 to 10 slides, one idea per slide, 6 to 12 words per slide. Slide 1 is the hook verbatim from the post body. Last slide is a real screenshot, not a CTA. Build at 1080x1350, export PDF, publish via /rest/documents.

The first 14 days, specific posts

ניהלתי מסעדה חמש שנים לפני שהתחלתי לכתוב קוד בשביל פרנסה. בגלל זה אני מוריד ללקוחות חצי מהפיצ׳רים שהם מבקשים.

כמה זה עולה לך היום, בשעות או בשקלים, בשבוע?

בניתי דמו של מערכת שבודקת מחירי מתחרים כל לילה. הנתונים בדמו מומצאים. המערכת האמיתית עוברת 1,121 מוצרים כל לילה — והמספר הזה אמיתי.

I gave five AI agents write access to my business. The interesting part is not what they do. It is the actions they cannot take without my signature.

רוב השיחות שמתחילות ב"אני צריך אפליקציה" נגמרות בזה שאני אומר שלא צריך אפליקציה. אלה ארבע השאלות שאני שואל לפני שאני מתמחר משהו.

אם על שלוש מהן התשובה היא "לא בדקתי" — אל תשלם עדיין לאף אחד שיבנה.

יש באתר שלי דוגמה של מסמך אפיון אמיתי, פתוח מהתחלה עד הסוף. לינק בתגובה.

מנהל רשת שמריץ תחרות בין סניפים בקובץ אקסל. זה המסך שמחליף את הקובץ.

Three forwardable assets, weeks 2, 4 and 5

Reach beyond a zero-follower network requires posts a stranger forwards to a colleague. That happens with a reusable artifact or a counter-consensus claim with receipts. He has both.

I built 10 working systems as clickable demos. Sample data, honestly labelled. Steal the ideas.

זה מסמך אמיתי, פתוח, אפשר לקרוא אותו מהתחלה עד הסוף לפני שמחליטים לשלם למישהו על אחד כזה.

Cadence, the floor

Do not go to 5 a week. The "5x posts equals 12x growth" claim is SEO noise. For a founder with two live builds it produces slop, and slop is unrecoverable for a technical seller.

Batching: one 90-minute block on Sunday morning produces the whole week's posts. Comments are daily and mobile. If Sunday is lost to a build, the post count drops to 1 and the comments do not drop at all. Comments are the load-bearing half.

Timing: Sunday to Wednesday, 08:30 to 10:00 IDT for the Israeli-buyer pillars. Tuesday to Thursday 09:00 IDT for English pillar-4 posts, which catches the Israel morning plus the European start of day. Never Friday, never Saturday.

Chagim adaptation, now until 4 October: shift to pillars 1, 3 and 4, none of which require a buying decision, and idea-stage people keep scrolling. Cut pillar 2 to roughly one post a month, since ops and finance managers are on vacation with kids. Do not attempt to close anything. Book for October.

The anti-slop pipeline: what agents do and what Tom does

Cost: about 12 to 15 minutes per post instead of about 40, and the words are literally his.

Mechanical enforcement, built into code rather than into good intentions

• Source gate. The publisher refuses to queue a post whose frontmatter lacks a source: field pointing at something real: a commit, a demo URL, a screenshot, a decision doc, a scrape run. The feed is then mechanically constrained to be a log of real work.
• Fact-check and blocklist agent (memory/feedback_linkedin_voice.md). Rejects any draft containing a client name, branch count, category, city or real SKU, any number without a repo source, and the 2026 slop fingerprint: one-sentence-per-line staccato throughout, "Here's the thing.", "Let that sink in.", "The result?", "It's not X. It's Y." as a closer, rhetorical-question openers, emoji bullets, "In today's fast-paced world", generic "3 lessons" lists, and premium / boutique / high-end.

The final test: Tom reads the post aloud in Hebrew. If he cringes, it isn't his and it doesn't ship.

Banned content types, encoded as a reject list

1. "AI will replace X" takes.
2. Tool listicles ("10 n8n nodes"). Attracts builders, and builders don't pay him.
3. Screenshots of his own ChatGPT or Claude conversations. The most recognisable slop signal in 2026, and it reads as "I have no work of my own to show." Inversion: a screenshot of his system's output, the WhatsApp morning brief or the Conductor console, is the opposite signal and is proof.
4. Anything implying a client relationship or outcome he can't evidence, and anything that lets a reader identify the retail client.
5. Polls, "agree?", "comment X for the link", follower milestones, "I'm humbled."
6. Day-in-the-life and productivity porn.
7. Reposts with a one-line comment.
8. Israeli politics or security commentary. The account is a business asset. Non-negotiable.
9. Any post before the profile rebuild is complete.
10. Content aimed at the מוסך/יבואן owner-operator. He is not here, and pretending otherwise corrupts the strategy for the audience LinkedIn actually delivers.

06 / Connections and DMs

Seed 40 guaranteed accepts before one cold request goes out

Phase 0, this week

Send 40 to 60 noteless invites, 10 to 15 a day, to people who accept at roughly 90% or better: former restaurant suppliers and staff, business-degree classmates, army contacts, the GoMobile people he already works with, Reey and the Monesys circle, and everyone in his phone who has a LinkedIn. Use My Network suggestions and the contact-sync import once, email only, never phone-contact sync, because that uploads his full address book including clients.

The 40-name comment list, which is the actual channel

Follow all 40, follow rather than connect. Notification bell on the top 10.

The comment standard: every comment is a mini-post

2 to 4 sentences, containing one specific thing the original post did not say: a number, a failure mode, what breaks at scale. Never "great post." Never a link. Never a pitch. A question only if it is genuinely open. Always in the language of the post, always within 90 minutes of publish, on posts that currently have 5 to 40 comments.

The piece missing from most agent setups is the approval gate. I run five agents on my own business and a defined set of their write actions cannot execute without my signature. Without that you do not have an operator, you have a liability.

The thing that bites at month three is silent failure. Ten scheduled jobs is fine. Ten scheduled jobs where only three alert on failure is how you find out in week six.

ההקלדה הכפולה בין שתי המערכות היא בדרך כלל החלק הזול לתקן. החלק הקשה הוא להחליט איזו מערכת היא מקור האמת. ההחלטה הזאת היא 80% מהפרויקט.

Targeting: hunt by intent, not by title

Search quota note: searching by name in the top box, and browsing his own 1st-degree connections, do not count against the commercial use limit, which is roughly 300 a month, undisclosed, resets at midnight PST on the 1st, and Support will not lift it. So McLead does target identification off-platform and hands over named lists, preserving his entire search quota.

Literal message sequences

היי [שם], ראיתי ש[חברה] עובדים על [X]. אני בונה מערכות אוטומציה לעסקים בארץ - בין השאר מערכת שבודקת מחירי מתחרים אצל לקוח קמעונאי, 1,121 מוצרים כל לילה. סקרן לראות מה אתם בונים.

תודה על החיבור. לא מוכר לך כלום - שאלה אמיתית: מה התהליך שהכי מעצבן אתכם היום, זה שמישהו מקליד ידנית משהו שמחשב היה צריך לעשות? אני עונה על זה בחינם, גם אם לא יוצא מזה כלום.

אהלן [שם], ראיתי מה שכתבת על [X] — הנקודה על [Y] מדויקת. אני בונה מערכות בדיוק בפינה הזאת ויש לי דמו שמדגים את זה עם דאטה מומצאת. רוצה שאשלח? בלי שום דבר מעבר לזה. תודה

Safe volumes and the ramp, for a dormant aged account

Profile views stay at 50 a day or fewer throughout.

Why volume follows phase rather than a fixed cap: an aged but dormant account is the profile LinkedIn throttles hardest in the first month, and there is nothing to gain from 60 invites a week while the recipients are answering "נדבר אחרי החגים." The ceiling opens when the profile is finished and the market is back.

Sales Navigator, first week of October and not before

Core is $119.99 a month, and the trial requires no LinkedIn free trial in the past 365 days, so burning it in August destroys the option for October.

What it unlocks: removal of the commercial use limit, lifting of the roughly 5 personalised notes per month cap, higher profile-view ceilings, 50 InMails a month. What it does not unlock: any API access whatsoever.

Decision rule, set in advance so it is not relitigated emotionally: if the trial month does not produce at least 2 real conversations, cancel.

07 / Measurement

Five numbers, checked weekly

Do not treat likes or impressions as success metrics before week 4. At near-zero followers, impressions are the platform's lottery, not a signal.

What McLead builds in AIOS

08 / Week 1 to week 4

The calendar

09 / Kill criteria and honest checkpoints

Four pre-committed decision points, so none of this gets relitigated emotionally

10 / The five specialist verdicts

What each lens concluded, including where they disagreed

11 / Closing

What only Tom can do, in order