Content Pack Review: Opus

Overall thoughts

• The em-dash rule is already clean. Every em-dash in the pack sits in a markdown header or file scaffolding (## Post 2 — The Friday backlog), never in publishable body copy. Nothing to fix there. The bigger AI tell in this pack is a different habit: the antithesis closer. "The fire is new. The discipline isn't." / "The ledger is the ground truth. The summary is a courtesy." / "you have a story, not a system" / "it was a demo. Not a system." / "Not demos. Working systems." That structure appears in nine places across five files. One is a signature. Nine is a template, and templates read as machine-written faster than punctuation does. My rewrites keep it in the About and Post 1 and strip it everywhere else, replacing it with a concrete instruction or a warm human aside.
• The teaching is real but thin in places. Posts 2, 3 and 5 genuinely teach. Posts 1, 4 and 6 assert. Post 4 in particular says a review queue "is the product" without saying how to build one that does not rot, which is the actual expertise. I have deepened those three: hashing and quarantine and table versioning in Post 3, ranking-by-money and one-screen and floor/ceiling in Post 4, the three grades with the "leave alone" criteria in Post 6.
• Half the batch is a numbered list. Posts 4, 5 and 6 all open with 1. 2. 3., and on LinkedIn three listicles out of six reads as content-marketing output regardless of how good the content is. Keep the list form for Post 5, where it is the point, and let the others run as prose with breathing room.
• Fintech and credit are missing entirely. Five of six posts are oilfield paperwork. The About names "fintech and credit companies" as a core vertical and it is also the owner's own deepest domain, which makes it the most defensible authority in the pack. A credit-memo or spreading post would be the single highest-value addition to batch 02, and it costs nothing in proof integrity because it is drawn from lived work rather than client stories.
• Nothing is shown, everything is told. The brand manual asks for a real labeled example. There is no post that walks a single document through the system field by field. That can be built from a synthetic field ticket with no invented client attached, and it would outperform every abstract post here.
• Proof integrity is intact. No invented clients, no fabricated ROI, no testimonials. The one place to stay alert is illustrative numbers: my Post 4 rewrite uses "ninety percent" and "sixty percent" as threshold examples, which is fine as an argument but should never migrate into a claim about results.

About standard

Thoughts: Both versions are on voice. The current one opens with a promise ("There is a version of your company where the paperwork handles itself"), which is the more commercial move; my earlier rewrite opens with an observation, which is the more composed one and matches the quiet posture better. Keep the Opus version, and note that it adds one line the current draft is missing: "Whether we build any of it is a separate conversation," which is the most disarming sentence available to a new advisory.

Current

There is a version of your company where the paperwork handles itself.

Hearth Advisory builds it. We design AI systems for the workflows traditional businesses run on: field tickets, invoices, credit memos, rate sheets. Not demos. Working systems your analysts, operators, and auditors can sign off on, because every answer shows the page it came from.

We work with oilfield service, industrial, and fintech/credit teams, where the paperwork is the product and a wrong number has consequences. That is exactly where careful automation pays for itself.

Every engagement starts with the AI Modernization Audit: fixed scope, fixed fee, and a written map of where automation returns hours to judgment and where it should never touch the process. You keep the map either way.

The fire is new. The discipline isn't.

Opus rewrite

The paperwork was never the point. It is what the work leaves behind: field tickets, invoices, credit memos, rate sheets.

Hearth Advisory builds the AI systems that carry it. Not demos. Working systems, with receipts: every answer shows the page it came from, so your analysts, operators, and auditors can put their name on it.

We work with oilfield service, industrial, and fintech and credit companies. Careful industries, where the paperwork is the product and a wrong number travels. That suits us.

Every engagement starts the same way: the AI Modernization Audit. Fixed scope, fixed fee, and a written map of where automation earns its keep and where it has no business going. You keep the map either way. Whether we build any of it is a separate conversation.

The fire is new. The discipline isn't.

About short

Thoughts: My earlier short variant is the right content but the wrong length for a tagline-adjacent slot, where it will be truncated. The version below tightens it and drops "industrial" so the three named verticals stay memorable. Honest note: the current short's "oilfield, service, and fintech" is vaguer than it needs to be, since "service" alone signals nothing.

Current

AI modernization for oilfield, service, and fintech companies. Working systems your team can sign off on, starting with a fixed-scope AI Modernization Audit. The fire is new. The discipline isn't.

Opus rewrite

AI modernization for oilfield service, industrial, and credit companies. Field tickets, invoices, credit memos, rate sheets: handled by systems your team can sign off on, because every answer shows the page it came from. Starts with a fixed-scope AI Modernization Audit. You keep the map either way.

Post 1: Quiet launch

Thoughts: Rewrite. "AI is the new fire" is the single most reused opener in AI thought leadership, and putting it in the first line of the launch post spends the brand's best idea on the reader's most skeptical second. It also teaches nothing, which breaks the rule that every post carries something reusable. Lead with the concrete rule, let the metaphor close.

Current

AI is the new fire. Enormous energy, dangerous loose, transformative held.

Most companies do not need more AI. They need the structure that holds it: systems their analysts, operators, and auditors can actually sign off on.

That is the work at Hearth Advisory. We modernize the workflows traditional businesses run on, from field tickets to credit memos, and every engagement starts with a fixed-scope AI Modernization Audit.

The fire is new. The discipline isn't.

Opus rewrite

Hearth Advisory is open.

We build AI systems for the paperwork traditional businesses run on: field tickets, invoices, credit memos, rate sheets.

One rule shapes everything we ship. Every answer shows the page it came from. Not a confidence score on its own, not a summary you have to take on faith. The document, the page, the line. If an analyst cannot check it in five seconds, it does not go in.

That rule costs something. It rules out a few popular architectures and it makes the first build slower than the demo you were shown elsewhere. It is also the reason the work survives an audit, a dispute, and a change of staff.

Every engagement starts with the AI Modernization Audit: fixed scope, fixed fee, and a written map of where automation earns its keep and where it has no business going.

The fire is new. The discipline isn't.

Post 2: The Friday backlog

Thoughts: The best post in the pack and the closest to the voice the brand is aiming for. My only complaint is that "the comparison" stays abstract for one beat too long. Naming the four checks turns a good observation into something an operations lead recognises as their own week.

Current

Every oilfield service company has a Friday backlog. Field tickets pile up all week, and someone spends Friday matching them to invoices by hand.

Here is where most automation projects go wrong: they automate the typing.

The typing is not the bottleneck. The bottleneck is the comparison. Does this ticket match this rate, this job, this PO? That is a judgment call repeated hundreds of times a week, and it is exactly the kind of judgment AI can draft for a human to confirm.

Automate the comparison. Keep the person on the exceptions. The Friday backlog disappears, and nobody's job had to.

Opus rewrite

Every oilfield service company has a Friday backlog. Tickets accumulate all week, and someone spends Friday matching them to invoices by hand.

Most automation projects aim at the typing. The typing was never the expensive part.

The expensive part is the comparison. Does this ticket match the rate on the contract, the scope on the PO, the hours on the crew's day sheet, the equipment actually on location? Four checks, a few hundred times a week, run by one person who is tired by two o'clock.

That is the piece worth automating. The system drafts each match and shows its work. Anything that agrees on all four passes through. Anything that disagrees goes to a human with the disagreement highlighted, not the whole document.

The coordinator stops reconciling every ticket and starts reviewing only the ones that did not line up.

Same headcount. Different Friday.

Post 3: The failure mode nobody demos

Thoughts: Excellent premise, and "archaeology" is the best word in the pack. The fix, though, is stated as two rules rather than taught as mechanics, and one of those rules now lives in Post 1. Push this one into the plumbing, where the credibility is.

Current

A vendor updates a rate sheet. Nobody tells the model.

The system does not start producing obviously wrong answers. It produces right-looking answers with old numbers. The invoice goes out. The dispute arrives three weeks later, and now a human is doing archaeology.

Two rules fix this. Every extraction cites the page it came from. And the system flags when a source document changes underneath it.

The ledger is the ground truth. The summary is a courtesy.

Opus rewrite

A vendor changes a rate sheet. Nobody tells the model.

This is the failure that does not look like failure. The system keeps answering. The answers keep looking right. They are simply built on a number that expired in March. The invoice goes out, the dispute arrives three weeks later, and someone is doing archaeology on their own billing.

Extraction accuracy will not catch this. The extraction was accurate. The source moved.

What catches it is unglamorous plumbing.

Hash every source document on ingest. When the hash changes, quarantine everything downstream of it and produce the list of invoices that touched the old version.

Version the rate table rather than overwriting it, so any past answer can be replayed against the table that was live on the day it was produced. That is what turns a dispute from an investigation into a lookup.

Ask a vendor how their system knows a source document changed. The length of the pause tells you most of what you need.

Post 4: How a review queue should actually work

Thoughts: Rewrite. This is the most valuable topic in the batch and the thinnest treatment of it. Three obvious bullets plus the claim that the queue "is the product" asserts expertise instead of demonstrating it. The real knowledge is why queues rot, and that is teachable in a post.

Current

1. Every extraction carries a confidence score and a link to its source page.
2. High confidence posts straight through. Low confidence lands in a queue.
3. The reviewer confirms or corrects in one tap. Corrections feed back into the system.

That queue is not a fallback. It is the product. It is where trust is earned, where auditors look first, and where the system proves it knows what it does not know.

Opus rewrite

The review queue is the part of an AI system nobody designs and everybody lives in.

Most get built as an afterthought: a table of low-confidence rows, no order, no owner, no end. It fills. People stop opening it. Six weeks later the queue is the same backlog you bought the system to delete.

Four properties separate a queue that holds up from one that rots.

It is ranked by exposure, not by confidence. A ninety percent match on a large invoice deserves a human before a sixty percent match on a small one. Sorting by confidence sorts by the machine's comfort instead of your risk.

It is one screen. Document on the left, extracted fields on the right, the cited page already open at the right line. If a reviewer has to go hunting in another system, the queue has quietly become a second job.

It has a floor and a ceiling. Too few items arriving means the thresholds are too loose and things are passing through unchecked. More items than one person can clear in an hour a day means the system is not ready for the next workflow yet.

Corrections are read, not just stored. Someone owns a monthly pass over them. Corrections are the only honest map of where the model is weak in your documents, in your vocabulary, on your worst vendor's formatting.

Build the queue first. The extraction is the easy half.

Post 5: Five questions that end an AI vendor demo early

Thoughts: Keep the spine, this is the most saveable thing in the batch. The one change worth making is adding a line under each question explaining what a bad answer sounds like. Bare questions posture. Questions with tells actually arm the reader.

Current

1. Show me a wrong answer. What does the system do next?
2. Where does this number come from? Show me the source page.
3. What happens when my vendor changes their rate sheet format?
4. Can my auditor reconstruct what happened, from the record, without calling me?
5. What does this cost per document at my volume, not yours?

If the demo cannot survive these, it was a demo. Not a system.

Opus rewrite

Five questions worth asking in an AI vendor demo. The answers matter less than how quickly they arrive.

Show me a wrong answer, and what the system did next. Every system is wrong sometimes. You are buying the recovery, not the accuracy.

Where did this number come from? Open the page. Time it. If it takes the vendor more than a few seconds with their own product, your auditor has no chance.

What happens when a vendor changes their rate sheet format? Listen for whether anything actively detects the change, or whether the honest answer is that you find out from a dispute.

Can my auditor reconstruct this from the record without calling you? "We can pull logs for you" is a no.

What does this cost per document at my volume? Ask for the number at your worst month, not your average one. Volume in this work arrives in spikes, and so do bills.

Good vendors enjoy these questions. They tend to have the answers ready, and usually a story about learning one of them the hard way.

Post 6: What an AI Modernization Audit actually is

Thoughts: Rewrite, lightly. The bones are right and the CTA is correctly understated. What is missing is the "leave alone" grade, which is the most distinctive and least copyable thing this offer does. Selling the refusal is what makes the rest credible.

Current

1. We map the workflows your business runs on. Not the org chart version. The real one, the one your coordinators could draw from memory.
2. We grade each one: automate now, automate with guardrails, or leave alone.
3. You get a written map, ranked by hours returned to judgment, with the guardrails specified.

Fixed scope. Fixed fee. No obligation to build anything. You keep the map either way.

If you want that map for your operation, message the Page.

Opus rewrite

What an AI Modernization Audit actually involves.

First we map the workflows the business really runs on. Not the org chart version. The version your coordinators could draw from memory, including the spreadsheet nobody officially supports and the person everyone calls when it breaks.

Then every workflow gets one of three grades.

Automate now. High volume, stable inputs, a clear right answer, and a cheap way to check the work. These pay for themselves quickly and they are where we start.

Automate with guardrails. Worth doing, but it does not go near production without a review queue, a citation trail, and a named owner. Most of the real value sits here, and so does most of the risk of doing it badly.

Leave alone. The reasoning is the value, or the volume is too low to matter, or the cost of being quietly wrong exceeds the hours saved. Audits usually find more of these than clients expect. Saying so plainly is a large part of what you are paying for.

You get the map ranked by hours returned to judgment, with the guardrails specified and the sequencing argued rather than asserted.

Fixed scope, fixed fee, no obligation to build anything. You keep the map either way.

If you want one for your operation, message the Page.

X Thread 1: The Friday backlog

Thoughts: Currently a verbatim copy of LinkedIn Post 2 with slash numbering, which is the least effective way to atomize. X rewards a first post that stands alone if nothing else is read, and shorter beats per tweet. Rewrite for the platform rather than reformatting for it.

Current

1/ Every oilfield service company has a Friday backlog. Field tickets pile up all week, and someone spends Friday matching them to invoices by hand. 2/ Most automation projects attack the typing. Wrong target. Typing is cheap. 3/ The bottleneck is the comparison. Does this ticket match this rate, this job, this PO? A judgment call repeated hundreds of times a week. 4/ That is exactly the judgment AI can draft and a human can confirm. Automate the comparison. Keep the person on the exceptions. 5/ The Friday backlog disappears. Nobody's job had to.

Opus rewrite

1/ Every oilfield service company has a Friday backlog. Tickets all week, one person matching them to invoices on Friday.

Most automation aims at the typing. Typing was never the expensive part.

2/ The expensive part is the comparison.

Ticket against the contract rate. Against the PO scope. Against the hours on the day sheet. Against the equipment actually on location.

3/ Four checks. A few hundred times a week. Run by someone who is tired by two o'clock.

4/ So automate the comparison, not the keystrokes.

The system drafts each match and shows its work. Agreement passes through. Disagreement goes to a human with the conflict highlighted, not the whole document.

5/ Same headcount. Different Friday.

X Thread 2: The failure mode nobody demos

Thoughts: Strongest thread of the three, and the premise is genuinely underdiscussed on X. Same note as the LinkedIn version: trade the aphorism ending for the mechanics, which is what gets bookmarked.

Current

1/ A vendor updates a rate sheet. Nobody tells the model. 2/ The system does not start producing obviously wrong answers. It produces right-looking answers with old numbers. 3/ The invoice goes out. The dispute arrives three weeks later. Now a human is doing archaeology. 4/ Two rules fix this. Every extraction cites the page it came from. And the system flags when a source document changes underneath it. 5/ The ledger is the ground truth. The summary is a courtesy.

Opus rewrite

1/ A vendor changes a rate sheet. Nobody tells the model.

This is the failure that does not look like failure.

2/ The system keeps answering. The answers keep looking right. They are built on a number that expired in March.

3/ Invoice goes out. Dispute arrives three weeks later. Someone is now doing archaeology on their own billing.

4/ Extraction accuracy will not catch this. The extraction was accurate. The source moved.

5/ The fix is unglamorous.

Hash every source document on ingest. When the hash changes, quarantine what depends on it and produce the list of invoices that used the old version.

6/ Version the rate table instead of overwriting it. Any past answer can then be replayed against the table that was live that day.

A dispute stops being an investigation and becomes a lookup.

7/ Ask a vendor how their system knows a source document changed.

The length of the pause tells you most of what you need.

X Thread 3: Five questions that end an AI vendor demo early

Thoughts: Weakest as a thread. Splitting five short questions across seven tweets gives each beat almost no substance, and the payoff is stranded in tweet 7 where most readers never arrive. Either run it as one dense post plus a quote-card image, or keep the thread and give every question a tell so each tweet earns its place. Rewrite below does the latter.

Current

1/ Five questions that end an AI vendor demo early: 2/ Show me a wrong answer. What does the system do next? 3/ Where does this number come from? Show me the source page. 4/ What happens when my vendor changes their rate sheet format? 5/ Can my auditor reconstruct what happened, from the record, without calling me? 6/ What does this cost per document at my volume, not yours? 7/ If the demo cannot survive these, it was a demo. Not a system.

Opus rewrite

1/ Five questions worth asking in an AI vendor demo.

The answers matter less than how quickly they arrive.

2/ "Show me a wrong answer, and what the system did next."

Every system is wrong sometimes. You are buying the recovery, not the accuracy.

3/ "Where did this number come from? Open the page."

Time it. If the vendor needs more than a few seconds inside their own product, your auditor has no chance.

4/ "What happens when a vendor changes their rate sheet format?"

Listen for whether anything detects the change, or whether the honest answer is that you find out from a dispute.

5/ "Can my auditor reconstruct this from the record without calling you?"

"We can pull logs for you" is a no.

6/ "What does this cost per document at my volume?"

Ask for the number at your worst month, not your average one. Volume arrives in spikes and so do bills.

7/ Good vendors enjoy these questions.

They usually have the answers ready, and a story about learning one of them the hard way.

X standalones

Thoughts: Two of the four are strong and reusable. The fire/hearth line is brand poetry with no takeaway and should live in the profile bio or a pinned reply, not the timeline, given the no-vague-inspiration rule. The 5%/95% line reads as a statistic and will be quoted back as one, so it is worth rephrasing as an argument instead of a number. I have added two replacements that carry actual mechanics.

Current

• A metric worth stealing: hours returned to judgment. Not tasks automated. Not accuracy. Hours your analysts get back for the work only they can do.
• AI strategy for most traditional companies is 5% choosing models and 95% finding the workflow where the paperwork hurts the most.
• The test for any back-office AI: could your auditor reconstruct what happened from the record, without calling you? If not, you have a story, not a system.
• Fire was never the point. The hearth was: the structure that made fire useful, safe, and worth gathering around. Same with AI.

Opus rewrite

• A metric worth stealing: hours returned to judgment. Not tasks automated. Not accuracy. Hours your analysts get back for the work only they can do.

• For most traditional companies, the model choice is the small decision. The large one is which workflow you touch first, and that should be chosen by where the paperwork hurts, not by where the demo looks best.

• The test for any back-office AI: could your auditor reconstruct what happened from the record, without calling you?

• If a system cannot tell you which page a number came from, it is not doing extraction. It is doing recall, and recall is not evidence.

• The cheapest guardrail in document AI is the refusal. A system that says "this rate table changed on Tuesday, I am holding the three invoices that depend on it" is worth more than one that is always confident.

• Sort your review queue by exposure, not by confidence. Confidence sorts by the model's comfort. Exposure sorts by yours.

(Retain the hearth line for the profile bio or a pinned post, where the metaphor is doing brand work rather than standing in for a takeaway.)

Content calendar

Thoughts: The posture is right: Page-first, twice a week, engagement week instead of volume, nothing publishes outside Buffer approval. Two structural problems. Six posts across three weeks followed by silence reads as a launch that ran out of material rather than a deliberate cadence, and the soft CTA lands in week 3 before there is enough published work to justify it. Spreading to 2/2/1/1 fixes both and holds the audit post until the proof posts have done their job.

Current

Week 1 — Arrival

• Tue: Post 1 (quiet launch) — pin it
• Thu: Post 2 (the Friday backlog)

Week 2 — Ground truth

• Tue: Post 3 (the failure mode nobody demos)
• Thu: Post 4 (how a review queue should work)

Week 3 — Proof discipline

• Tue: Post 5 (five questions that end an AI vendor demo)
• Thu: Post 6 (what the audit actually is, soft CTA)

Week 4 — Engagement, not broadcasting

• No new posts. Comment thoughtfully on 3–5 target posts/week from the Page. Drafts for each comment come from me first.
• Review: what got impressions, what got profile visits, what felt off. Decide batch 02 from evidence.

Opus rewrite

Week 1: Arrival - Tue: Post 1 (quiet launch). Pin it. - Thu: Post 2 (the Friday backlog) - Engagement floor starts now: 3 substantive comments per week from the Page, never a pitch.

Week 2: Ground truth - Tue: Post 3 (the failure mode nobody demos) - Thu: Post 4 (how a review queue should actually work)

Week 3: Proof discipline - Tue: Post 5 (five questions that end an AI vendor demo) - Thu: engagement only. If any of Posts 1 to 5 drew a real question in the comments, the answer to that question becomes the Thursday post instead. A reactive post beats a scheduled one every time.

Week 4: The offer, quietly - Tue: Post 6 (what the audit actually is, soft CTA). It lands after five posts of evidence rather than in the middle of them. - Thu: engagement only. - End of week: review and decide batch 02 from evidence.

What to actually measure in the week 4 review

Impressions are the least useful number available at this stage. Rank by, in order: 1. Profile and Page visits per post. This is intent. 2. Saves and sends. This is whether the teaching landed. 3. Comments from people who hold the job you are describing, as opposed to peers in consulting. 4. Inbound messages, however few. One from an operations lead outweighs a thousand impressions.

Batch 02 priorities, in order

1. A credit or fintech post. The vertical is named in the About and appears nowhere in batch 01, and it is the owner's strongest personal ground. Credit memo spreading, covenant tracking, or exception queues in a lending workflow.
2. A shown example. One synthetic field ticket or rate sheet, walked field by field, with the citation and the confidence and the queue decision visible. No client attached, no metrics claimed. This is the post that would outperform everything in batch 01.
3. A "what we told a client not to automate" post, written generically enough to require no client. The refusal is the most distinctive thing in the offer and it currently exists only as a bullet inside Post 6.
4. Audit FAQ follow-ups drawn from actual week 1 to 4 replies, not anticipated ones.

Keep unchanged: the Buffer draft-and-approve workflow, X staying dark until separately activated, the no-pitching-in-comments rule, and the engagement target list.

The tagline question

Recommendation: AI modernization for oilfield service, industrial, and credit companies. Every answer shows its source page.

Why: The tagline slot is not a brand slot, it is a search-and-context slot: it appears under the Page name in results, in the byline of every post, and beside every comment left from the Page, which means it is read by people who are deciding in one second whether this is their world. "Keep the fire." asks a stranger to already know the brand in order to decode it, and for a Page with no audience that spends the strongest line in the system on the least receptive moment. Descriptive wins here, but the current descriptive underperforms for its own reasons: "oilfield, service, and fintech" reads as three loose words, and "service" on its own signals nothing, so the fix is to name the verticals precisely and then earn the rest of the character budget. The second sentence does the differentiating work at 108 characters, front-loaded so the qualifying clause survives any truncation, and it converts the offer from a category claim into the one mechanic nobody else puts in a tagline. Poetry is right for surfaces where the reader has already arrived; clarity is right for the surface that decides whether they do.

Where the other line goes: "Keep the fire." belongs on the cover image, which the brand manual already specifies as a one-line serif tagline in ivory with the ember dot as its period. That is the correct home: it is seen only after someone has landed on the Page, where the metaphor is doing brand work rather than standing in for an explanation. Keep it out of the About body, which already closes on "The fire is new. The discipline isn't." and does not need two fire lines, and keep it out of post copy entirely. Its second legitimate use is a sign-off on ceremonial surfaces only: the founder letter, a milestone announcement, print and merchandise.

If you want options: - Poetic: Keep the fire. AI modernization for oilfield service, industrial, and credit companies. (86 chars, leads with the brand line and still qualifies the reader) - Descriptive: AI modernization for oilfield service, industrial, and credit companies. (73 chars, the safest version, and the one to use if the second sentence ever reads as crowded in-feed)

The tagline question, round two

Fair point? Yes, on the wording. I spent the budget defending the instinct that the second clause should say the one thing only Hearth would say, and that part I still hold: naming the verticals qualifies the reader, but it does not differentiate you from every other consultancy that names the same three. Where I was wrong is in assuming the differentiator had to be stated as a mechanic. "Every answer shows its source page" is a spec line, and a spec only lands on someone who already feels the problem it solves. A stranger scrolling past a comment byline has not yet thought about citation failure, so the clause asks them to do work instead of making them feel something.

Revised recommendation: AI modernization for oilfield service, industrial, and credit companies. Numbers your team can put their name on.

Why this one pulls: It is the same claim, moved from the machine to the person. Nobody has ever felt anything about a source page, but everyone in a careful industry knows the specific weight of signing off on a number they did not personally verify, and that sentence names the feeling without explaining it. It reads as a promise rather than a feature, which is what a byline should be doing in the second someone gives it. The verticals stay in front, so the search value and the truncation behavior are unchanged at 114 characters. It also comes straight out of the About section, which already says "put their name on it," so the tagline now reads as the compressed version of the page rather than a separate piece of copy competing with it.

Alternates: - Descriptive: AI modernization for oilfield service, industrial, and credit companies. Built for the people who sign off. (106 chars. Flatter, more literal about who the buyer is. Use this if "numbers" ever reads as too narrow for the industrial work.) - Magnetic: AI modernization for oilfield service, industrial, and credit companies. The paperwork was never the point. (106 chars. Opens a small loop and makes the reader want the rest, and it is the first line of the About, so the payoff is one click away. Slightly riskier: it withholds where the recommended line delivers.)

The discovery filing

Thoughts: The industry pick is right and the credit depth does not change it, but roughly a third of the specialty list is spending slots on words that describe Hearth rather than words a buyer types, and the credit vertical, which is the deepest personal expertise in the business and the flagship service page in the CreditIntel direction brief, is represented by two phrases nobody searches while "portfolio risk," "credit analysis," and "commercial lending" are missing entirely. The three hashtags are the weakest part: featured hashtags are an engagement surface, not a search surface, so the question is not which words describe the firm but which feeds the Page will actually comment in every week, and #Fintech is a feed of consumer payments and crypto news where no credit manager is waiting. The largest lead problem in the filing is not in any of the three fields being reviewed: the Page has no website, no CTA button, no location, and Post 6 closes with "message the Page" when Page messaging is off by default, so a buyer who is fully convinced has nowhere to go.

Current

Industry: Business Consulting and Services

Specialties (as filed, 20): AI modernization, workflow automation, intelligent document processing, field ticket automation, invoice automation, credit memo automation, accounts payable automation, back office automation, oilfield services, oil and gas technology, fintech, credit risk operations, process audit, AI readiness assessment, human-in-the-loop AI, operations consulting, digital transformation, document extraction, AI governance, modernization audit

Featured hashtags: #WorkflowAutomation #OilfieldServices #Fintech

Opus rewrite

Industry: Business Consulting and Services (unchanged)

Specialties, in order:

1. AI consulting
2. workflow automation
3. intelligent document processing
4. business process automation
5. invoice automation
6. accounts payable automation
7. credit risk management
8. credit analysis
9. portfolio risk management
10. commercial lending
11. credit memo automation
12. accounts receivable automation
13. field ticket automation
14. oilfield services automation
15. document automation
16. financial spreading
17. AI readiness assessment
18. operations consulting
19. back office automation
20. AI governance

Featured hashtags: #CreditRisk #OilfieldServices #WorkflowAutomation

What changed and why

• Industry stays. Financial Services would imply Hearth is a lender and invites a regulatory read the guardrails already refuse; IT Services and IT Consulting files the firm next to offshore dev shops and prices it like one; Oil and Gas buries it among equipment vendors. The credit depth belongs in the specialties and the flagship service page, not the category. If CreditIntel ever splits into its own Page, that one files under Financial Services and this one does not change.
• Cut "AI modernization" and "modernization audit" from slot one. Both are Hearth's own coinages with no query volume behind them. They are the right words for the tagline and the About, where a human reads them, and the wrong words for a matching field, where a machine does. The offer name loses nothing: it is stated twice in the About already.
• "AI consulting" takes slot one. It is the single highest-intent phrase a buyer types when they have budget and no vendor, and it is a plain category label rather than agency hype, so it brushes the avoid-list without crossing it. This is a machine-matching field rendered as a comma list in the Overview tab, so the brand cost is close to zero.
• Cut "fintech" and "oil and gas technology." Single broad vertical nouns match searches by people looking to buy from that industry, not from a consultant serving it, and a new Page ranks nowhere against the actual industry pages. "Oilfield services" becomes "oilfield services automation" for the same reason: the intersection is what Hearth wins, the bare vertical is what it loses.
• Five credit terms replace two weak ones. "Credit risk operations" and "process audit" are phrases from an internal document, not a search bar. Credit risk management, credit analysis, portfolio risk management, commercial lending, and financial spreading are what the named buyers in the CreditIntel brief actually type, and they cover both sides of the vertical: trade credit and lending.
• "Credit memo automation" and "field ticket automation" survive on the opposite logic. Almost nobody searches them, but almost nobody claims them either, so the few who do search land on a Page with no competition and perfect intent. That trade is worth two mid slots; it is not worth a top-five slot.
• Added "accounts receivable automation." The direction brief names AR and collections leaders as primary buyers and the filing had no term for them at all.
• Added "business process automation." A real procurement category with real volume, and it captures the operations buyer who has not yet framed the problem as an AI problem.
• Cut "human-in-the-loop AI" and "document extraction." The first is practitioner vocabulary that buyers do not search, and it is a message, which belongs in the posts where it already lives. The second is redundant against intelligent document processing plus document automation.
• Cut "digital transformation." It is the definition of a term that sounds impressive, it is the most crowded phrase in consulting, and the inbound it does pull is tire-kickers. Keep it as the first alternate if a slot opens: swap it for "financial spreading" if claiming spreading feels ahead of what has actually been built.
• Hashtags: keep #OilfieldServices, keep #WorkflowAutomation, drop #Fintech for #CreditRisk. The oilfield feed is mid-sized and full of actual operators, which is exactly where a Page with no audience can be visible. #Fintech is enormous and its content is consumer payments, crypto, and funding news, with no overlap with a credit manager's day. #CreditRisk is where the buyer described in the CreditIntel brief actually reads. Swap to #CreditManagement instead if the trade-credit and AR side is the priority over lending, and to #OilAndGas if the oilfield feed turns out too thin to comment in weekly.
• Choose hashtags by where the Page will comment, not by what the Page is. Featured hashtags mainly give the Page a feed to participate in and a small link row on the Page. They pay off only against the engagement floor already in the calendar, three substantive comments a week from the Page. If that habit does not happen, all three slots are decoration regardless of which words fill them.
• On ordering: LinkedIn does not publish how specialty order is weighted, so the honest reason to order carefully is the one that is verifiable: the Overview tab truncates, and a human skimming reads the first few. Ordering as though weight follows position costs nothing and is right for the reader either way.

Beyond the three fields, what is actually costing leads

• No website and no CTA button is the biggest hole in the filing. A buyer convinced by the About has no next step. Both can be fixed before the public site exists: point Website at a single approved landing page or even the booking link, and set the CTA to "Contact us" or "Book a walkthrough." An unfinished Page also reads as an inactive company to the exact careful buyer this brand is chasing.
• Turn on Page messaging or change Post 6's close. Post 6 ends with "message the Page," and Page messaging is admin-enabled, off by default. As filed, the only CTA in the entire content pack lands on a door that does not open.
• Location is missing entirely. A Page with no location loses every geography-qualified search, and the oilfield vertical is geographically concentrated. Add HQ city and state, plus founded year. Both are small credibility signals that cost nothing and are true.
• Ibrahim's personal profile is the stronger discovery asset and it is not in this filing. New Company Pages rank poorly and have no network; his profile has both. Listing Hearth Advisory as current position links the two, and LinkedIn's service-provider search runs off personal profiles rather than Pages, so the Services section on his profile is a lead channel the Page structurally cannot access.
• Size 0-1 employees will cost some enterprise credibility. It is also true, so keep it. The offset is proof content and the fixed-scope audit, not a different number.
• The About SEO note holds up, with one gap. All four keyword nouns land inside the first 156 characters of the Opus About: field tickets, invoices, credit memos, and rate sheets appear by character 120. What does not appear in the snippet is "AI" or "automation," which the 114-character tagline sitting directly under the Page name carries instead. That is a fair division of labor and not worth rewriting the opening for.

The services section

Thoughts: The blocker is real and it is not a wording problem. The Page services picker is a closed taxonomy, and LinkedIn has no service named Artificial Intelligence, Machine Learning, Data Analytics, Automation, Process Consulting, Operations Consulting, or Risk Management. Every one of those is a category the industry talks in and LinkedIn's marketplace does not. The taxonomy that does exist is organized around sixteen parent categories, and the ten slots have to be spent inside it. That is survivable: the strongest matches available are actually more concrete than the words we wanted, and one of them, Invoice Processing, describes the work more literally than anything in the original list. Where a term has no home, it moves to Specialties and to the About text, which are both freeform and both read by humans rather than by the matching engine.

How the picker actually behaves: Services are chosen from LinkedIn's predefined list, either by browsing the parent categories or by typing into the search box, which is the faster path. Ten maximum. One pricing setting applies to all ten; there is no per-service price. Work location defaults to the Page's primary location, with a separate checkbox for remote.

Services to select, in order

1. Business Consulting (Consulting) - The audit is the front door and this is the widest-intent consulting label a buyer types when they have a problem and no vendor. It carries the AI Modernization Audit better than anything else in the taxonomy.
2. Management Consulting (Consulting) - The other half of that same search. It is also LinkedIn's default routing category for cross-industry consulting, which matters on a Page with no history.
3. Invoice Processing (Operations) - The most literal match in the entire taxonomy to what Hearth actually does. Almost no consultancy claims it, so the few requests routed here arrive with perfect intent.
4. Business Analytics (Software Development) - The closest live proxy for the AI and data build work. It sits under Software Development rather than Consulting, which quietly signals that Hearth ships systems rather than decks.
5. Financial Consulting (Consulting) - Carries the credit and controller buyer inside the consulting category, where the audit buyer is already looking.
6. Financial Analysis (Finance) - Credit memos and spreading, in the term a controller or credit manager actually types.
7. Information Management (Software Development) - The nearest taxonomy home for intelligent document processing. Rate sheets, tickets, and memos as records that have to stay reconcilable.
8. Custom Software Development (Software Development) - Declares that the engagement can end in a working system. Without it, all nine other slots read as advisory only.
9. Data Reporting (Operations) - The internal web artifacts and the on-demand decision surfaces. This is what the deliverable looks like when it lands on someone's screen.
10. Enterprise Content Management (Software Development) - The document backbone term. It skews slightly legacy, but it is the only slot in the list that says the documents themselves are the system of record.

Bench, with swap conditions: Change Management (Coaching and Mentoring) if adoption and review-queue work becomes the lead story, though it routes through a coaching-heavy category. Strategic Planning or Program Management (Operations) if the operations buyer outnumbers the finance buyer. Financial Reporting (Finance) if the credit side outruns the oilfield side. Commercial Lending exists but files under Real Estate and would misread Hearth as a lender or broker, so leave it. IT Consulting exists and is tempting, and it is the same trap the industry field already avoided: it files the firm next to offshore dev shops and gets priced like one.

Desired services with no taxonomy match, and the substitute:

• Artificial Intelligence, Machine Learning, AI Consulting - No match anywhere in the picker. Covered by Business Analytics plus Custom Software Development. The word AI has to be carried by the services About text, the Specialties list, and the tagline instead.
• Data Analytics - No match. Business Analytics is the taxonomy's version of it.
• Automation, Workflow Automation, Business Process Automation - No match. Invoice Processing covers the concrete case and Business Consulting covers the advisory frame. Outsourcing exists and is the wrong signal entirely, so avoid it.
• Process Consulting, Operations Consulting - Neither exists under Consulting. Management Consulting is the substitute, with Strategic Planning under Operations as the alternate.
• Risk Management - No match. Financial Consulting and Financial Analysis carry it.
• Credit Repair and credit services - No match, and the near neighbors are consumer-credit terms that would attract exactly the wrong inbound. Financial Consulting is the substitute.
• Intelligent document processing, document extraction - No match. Information Management plus Enterprise Content Management.
• AI readiness assessment, AI governance - No match. Business Consulting.

Specialties: freeform, so the list stands

Specialties is still a free-text field in 2026, not a picker. You add them one at a time and type the term, up to twenty. The proof is already in the filing: "AI modernization" and "modernization audit" are Hearth's own coinages and they saved successfully, which no controlled vocabulary would have allowed. The twenty-term list from the discovery filing stands exactly as written, unchanged. This is also where every term that lost its picker fight goes, which is why the list already leads with AI consulting and carries workflow automation, intelligent document processing, and credit risk management. The two fields are doing different jobs: Specialties is read by search and by a human skimming the Overview tab, and Services is read by a routing engine with a fixed vocabulary.

The disclosure line

The work is safe to describe; the employer is not, and the two come apart cleanly because nothing in the track record needs a name to be persuasive. What can be said publicly right now: automated credit deep-dives and credit file filling, a full-portfolio rescreen with on-demand credit decisions that reads historical credit and collection notes into soft and hard signals, and delivery as internal web artifacts. What cannot: the employer's name, its industry position, any client or counterparty, the borrowing basis work still in flight, and any first-person sentence that ties a named person to a named company while the posture is quiet.

The exact phrasing rule for anonymized scale claims: state scale as a rounded order-of-magnitude band attached to an unnamed category noun, never as a figure. "A nine-figure monthly credit portfolio," not "$250M/month." "A multibillion-dollar payments company," not "$3B." Exact figures plus a named vertical are a fingerprint, and anyone inside payments can identify a $3B private company running $250M a month in about one guess. Two supporting rules: use one scale attribute per sentence rather than stacking company size and portfolio size together, because two bands triangulate faster than one, and always frame the work as in-house rather than delivered, so no reader can mistake an employer for a Hearth client. The safest construction is the portfolio band without the company band, since the portfolio is what proves the competence and the company descriptor only adds identifying surface.

About (500 characters max)

The form asks for services, what makes you stand out, and previous projects or years of experience. This covers all three in that order, in first person plural, with the scale claim banded per the rule above.

The paperwork was never the point. We build the AI systems that carry it: field tickets, invoices, credit memos, rate sheets. Every answer shows the page it came from, so your team can put their name on it. The discipline comes from in-house work on a nine-figure monthly credit portfolio: automated credit deep-dives, and a full-portfolio rescreen with on-demand decisions. Every engagement starts with a fixed-scope AI Modernization Audit. You keep the map either way.

470 characters. Thirty in reserve, which is deliberate: LinkedIn counts line breaks and occasionally trailing whitespace against the limit, and a field that saves at 499 will fail on the next small edit. "In-house work" is the load-bearing phrase. It says the track record is real and it forecloses the reading that a multibillion-dollar company was a Hearth client, which is the one claim in this section that would cost more than it earns.

Work location and pricing

Houston: correct as checked. The oilfield buyer is geographically concentrated and this is the only field on the Page that wins a geography-qualified search.

Remote: check it as well. Requests are matched on location, so leaving it unchecked silently drops every credit ops, controller, and finance buyer outside Houston, which is most of them. The delivery format is internal web artifacts, so the work has no physical dependency, and Houston stays as the anchor either way.

Pricing: select "Contact for pricing," not a starting hourly rate. The single rate applies to all ten services at once, so any number priced for the audit misprices the build and any number priced for the build scares off the audit. It also contradicts the offer directly: the audit is fixed scope and fixed fee, and publishing an hourly floor invites the buyer to negotiate the rate before they have understood the scope. "Contact for pricing" costs one message and keeps the first conversation about the work.

The growth strategy

The honest frame

LinkedIn will not produce the first audit. For a Page with no followers, no network, and a niche buyer, organic reach is close to zero and stays there until something outside the feed sends people in. What LinkedIn can do, starting now, is be the surface that survives the check: when an operations lead or a credit manager hears the name from a conversation, a referral, or an email, the Page is where they decide whether this is a real firm or a side project with a logo. In this market the first three engagements come from people Ibrahim can already reach directly, and the Page's job is to make that reach convert rather than to replace it. Everything below is built on that division of labor. Treat any post-level lead in the first 90 days as a bonus, not the plan.

Phase 0: Close the holes (week 0)

Nothing in Phase 1 or 2 is worth an hour until these are true. All of it is one focused afternoon.

1. Turn on Page messaging. It is admin-enabled and off by default. Post 6 closes with "message the Page," and as filed that CTA lands on a door that does not open. Turn on message notifications to the phone at the same time, because a 48-hour reply to the first inbound is worse than no CTA.
2. Decide the destination, and it has to be a real one-page site on Hearth's own domain. There is no public website, and the Website field is the only place on the Page that a convinced buyer can go. The minimum acceptable version is six blocks and under 700 words: what Hearth does in one sentence, who it is for by named vertical, the problem stated in the buyer's words, the audit block (five days, fixed scope, fixed fee, what you receive, "you keep the map either way"), the one rule (every answer shows the page it came from), and one action. No blog, no team page, no case studies, no pricing table, no chat widget. The raw material already exists: pick the Fable or K3 build under /opt/data/content/hearth/website-builds/ and strip it to a single page rather than waiting for the full site. If the domain genuinely cannot be live this week, a hosted single page is acceptable as a bridge only if it sits on Hearth's domain. A builder subdomain in the Website field reads as a hobby to exactly the careful buyer this brand wants. Never point the field at a Google Doc, a bare booking link, a Tailnet preview URL, or nothing.
3. Set the CTA button to "Contact us" pointing at the contact anchor on that page. "Learn more" is the weakest option available and "Sign up" misdescribes the offer.
4. Fill the rest of the credibility surface: Houston as location, founded year, the cover image with "Keep the fire." on it, logo, the twenty specialties, the ten services. A Page with empty fields reads as an inactive company.
5. Get an email on the domain. A gmail address in the contact block undoes the rest of the page.
6. Set the UTM convention before the first post goes out. ?utm_source=linkedin&utm_medium=page on the Website field and CTA button, utm_medium=post on in-post links, utm_medium=comment on anything dropped in a comment. Without this, "webpage visits" is a number nobody can act on, and the whole measurement section below stops working. Put lightweight analytics on the page the same afternoon.

Phase 1: Authority (weeks 1 to 6)

The content engine is already most of the way there. Three changes.

Cadence. Run the revised 2/2/1/1 calendar from the section above, which holds Post 6 until five posts of evidence have landed. From week 5 the sustainable shape is one scheduled post plus one reactive post per week, where the reactive one answers a real question someone asked in comments or in the inbox. A reactive post beats a scheduled one every time, and it costs less to write.

Hours. Against a 3 to 5 hour budget: 60 to 90 minutes writing, 90 minutes commenting, 30 minutes on measurement and inbox, the rest as slack. If a week compresses, cut the post and keep the comments. That priority order is deliberate and it is the opposite of what most people do.

Formats that earn a follow in this niche. Four work here: the mechanism post that shows the plumbing (Post 3), the diagnostic list that arms the reader for a decision they are actually facing (Post 5), the refusal post that names what should not be automated, and the shown artifact. Nothing else does. Skip commentary on model releases, funding news, and anything that begins with a statistic about AI adoption. That content attracts other consultants, and other consultants never buy.

What batch 02 must add, in order:

1. The credit post. The vertical is named in the About, it is Ibrahim's deepest ground, and it appears nowhere in batch 01. Credit memo spreading, covenant tracking, or the exception queue in a lending workflow. Written from in-house experience under the banding rule already established in the services section. This is the least contested authority Hearth has and it is currently invisible.
2. The shown-not-told document walkthrough. One synthetic field ticket and, separately, one synthetic credit memo, walked field by field: what was extracted, which page it cites, what confidence it carried, what the queue did with it, and the one field the system correctly refused to answer. Publish as a document post, not a text post, because the format holds attention longer and the artifact is the point. This is the single highest-value asset in the plan and the one that makes every abstract post retroactively credible. No client attached, no metrics claimed.
3. The refusal post. "What we tell people not to automate." The leave-alone grade currently exists as one bullet inside Post 6, and it is the most distinctive and least copyable thing in the offer.
4. The process transparency post. What actually happens across the five days of an audit, day by day. A $1,000 to $3,000 purchase from an unknown firm is bought on process clarity, not on capability claims.
5. Audit FAQ posts drawn from real replies, not anticipated ones.

Pin Post 1 through week 4, then swap the pin to the walkthrough once it exists, because that is the post that does the selling.

Phase 2: Distribution (weeks 2 to 8, overlapping)

Posting is not distribution. For this Page, comments are.

A substantive comment under a post that 300 relevant operators are reading reaches more of the right people than a Page post reaches in a month. The comment carries the Page name, the tagline, and one click to the Page, which is precisely the funnel Phase 3 needs. Everything else in this phase is secondary.

Target: five substantive comments per week, comment as the Page where LinkedIn permits it. Each one adds a mechanism the original post left out, or a specific question a practitioner would ask. Fifteen minutes each. If a comment could be posted under any other consultant's name, it is not worth leaving.

Named habitats. Build a list of roughly twenty accounts and feeds across three groups, and verify each is active before committing a slot: the oilfield operations group (the #OilfieldServices and #OilAndGas feeds, energy trade publications, and the Page feeds of ticketing and field-data vendors such as the OpenInvoice and OpenTicket ecosystem, WellSight, RigER, and OFS Portal, where customers air real pain in the comments); the credit and lending group (#CreditRisk and #CreditManagement, NACM and Credit Research Foundation content, commercial credit and AR leaders, lending ops voices in fintech); and the AP/AR automation vendor group, where the comment sections are full of controllers describing exactly the failure modes Hearth writes about. Vendor Page comment sections are the most underrated of the three because the audience is pre-qualified by frustration.

Never: engagement pods, comment automation, mass connection requests, follow-for-follow, bought followers, and above all cold DM pitching. In a buyer set this small and this connected, one pitchy message to an operations director costs more than a hundred followers gain.

The personal profile tension, stated plainly. Pre-disclosure, Ibrahim's profile is a strong asset that has to sit idle, and there is no clever way around that. Safe now: engaging with industry content under his own name as a practitioner, which is ordinary professional behavior. Not safe now: listing Hearth Advisory as a current position, announcing anything, reposting Page content, linking the two in a featured section, offering services, or building an audience in a way that reads as running something on the side. Also worth naming: the 250-per-month connection invites to follow the Page are the largest free lever LinkedIn offers a new Page, and they generate a notification with Ibrahim's name on it. They stay unspent until disclosure. That is a real cost of the quiet posture, and it should be booked as one rather than forgotten.

What changes when disclosure is safe. Sequence it over two to three weeks rather than one day: list Hearth as current position, which gives the Page access to his network's discovery surface for the first time; spend the invite credits in batches; turn on the Services section on his personal profile, because LinkedIn's service-provider search runs off personal profiles and not Pages, which makes it a lead channel the Page structurally cannot reach; begin posting as himself, since personal posts consistently outreach Page posts by a wide margin and the Page becomes the archive rather than the megaphone; and publish the founder letter, which is the one place the "Keep the fire." line earns its keep. Expect the growth curve to inflect here and not before. Do not read pre-disclosure numbers as a verdict on the content.

Phase 3: Conversion (ongoing)

The full path, with the ask at each step:

Comment (no CTA, ever; the byline and tagline are the CTA) → Page visit (About closes on the audit; the pinned post carries the proof) → Website click via the CTA button → one-page site (one action: email or a short form, and the audit block above the fold on mobile) → first message → 20-minute call → written audit proposal, one page, fixed fee, fixed dates → audit → build conversation, separately.

Handling "message the Page." The Page inbox is not a real inbox and it does not behave like one, so the habit has to be manual: check it daily, reply inside twelve hours. The first reply is not a pitch and never carries a deck. It is three questions: which document is the problem, roughly what volume per month, and who touches it today. Then propose twenty minutes. The audit is the ask on that call, not a proposal to send later.

When to turn on a booking link. Not at week 0. A booking link on a Page with no inbound is an ornament, it skips the qualification the three questions provide, and a calendar showing only evening slots quietly discloses that this is a side venture. Add it after the third real inbound conversation, when scheduling friction has become an actual cost. Until then, message and email convert better because they let Ibrahim control the first reply.

The half that is not LinkedIn. Inbound alone will not fill the first quarter. Run five to ten direct conversations a month through warm channels in parallel: former colleagues, industry contacts, referral asks. The Page is what those conversations check afterward, and that is the mechanism by which content actually generates revenue at this stage. If the Page work is happening and the direct conversations are not, the strategy is not running.

Measurement

Weekly, in this priority order:

1. Qualified conversations started. Inbound messages plus comment replies from someone who holds the job being described. One per month in the first sixty days is a healthy signal. This is the only number that connects to revenue.
2. Website clicks from LinkedIn, split by UTM medium. Comment-driven clicks versus post-driven clicks tells you which half of the engine is working.
3. Page unique visitors, and the visitor demographics breakdown. The breakdown matters more than the count. If visitors skew to consultants, students, and agency owners, the content is winning the wrong room and the habitat list needs changing.
4. Comment quality ratio: comments from job-holders versus peers.
5. Followers. Last, and deliberately so. It is the number most likely to be watched and the least predictive of anything.

Diagnostic thresholds. By day 60 there should be roughly twelve posts and forty comments behind you. Read the funnel by where it breaks: comments producing no Page visits means the habitats are wrong, not the writing; Page visits producing no site clicks means the About or the CTA button is failing; site clicks producing no conversations means the landing page is not making the audit concrete enough.

By day 90, working looks like two or more qualified conversations and one audit sold or scheduled. Change approach if the volume targets were genuinely met and the result is zero conversations: that is evidence LinkedIn is not the pre-disclosure channel, and the correct response is to drop the Page to one post per week on maintenance, move those hours into direct outreach and referrals, and reopen the full plan when disclosure lands. That is a legitimate outcome, not a failure, and deciding it on evidence at 90 days beats drifting for a year.

The one thing that matters most

Five substantive comments a week, in the named buyer feeds, sustained for eight weeks without a single pitch among them. Phase 0 is a chore to be finished once, and posts without distribution are a diary. The comment habit is the only mechanism in this plan that puts a zero-audience Page in front of the people who buy, it costs ninety minutes a week, and it is the first thing that gets dropped in a busy week. Protect it ahead of the posts.