Six outreach drafts Serge shared today for a reaction, covering three ICP segments, one lead magnet invite, one reconnect email and one screenshot-led message. Scored against the same seven-point diagnostic used on 67 previous teardowns. Two of the six are already close to right. The other four share one root cause.
6messages analysed
4.6avg score before
9.1avg score after
5 of 6closed on guest language
What the set reveals
Read on their own, these six read as reasonable cold outreach. Read as a set, a pattern shows up that no single message would surface: three of the six, aimed at three different buyer personas, are the same skeleton with the nouns swapped. Open with a diagnostic question about the reader's world, pivot on line three to "as part of building Itelgie, we're speaking with X," close with "would you be open to a 20-minute conversation."
That pivot is the expensive part. The question in line one belongs to the reader. The sentence right after it hands the email back to Serge's company. A reader who gets two of these three, on two different lists inside the same organisation, will recognise the shape before finishing the first sentence.
The two messages that break this pattern, the lead magnet invite and the screenshot message, are also the two highest-scoring in the set. That is not a coincidence. Both open with something concrete instead of a research framing, and both are structurally impossible to confuse with a template.
3/6Identical skeleton across three ICP segmentsFinTech/Healthtech, CEO/COO and HRD messages open the same way, pivot the same way and close the same way. Only the persona noun changes.
4/6Company-first pivotLine one asks about the reader. Line three switches to "as part of building Itelgie, an AI-powered platform for X, we're speaking with Y." The reader's problem is dropped the moment the product enters.
5/6Guest language close"Would you be open to a short conversation" and "Happy to connect" describe Serge's hope, not a decision the reader makes. Only the screenshot message ends on a question the reader actually owns.
5/6Zero named proofNo client name, no completed pilot, no number tied to a real account anywhere in the set. The one statistic that appears, 30 to 50 percent faster time-to-hire, is unsourced and not tied to Itelgie itself.
3/6Unbuilt incentive offered as fact"We'll share a benchmarking report with all participants" appears three times. Nothing in the thread suggests this report exists yet. A COO who asks to see it before replying has no answer waiting.
1/6Real personalisationOnly the reconnect email references something specific about the recipient, the earlier conversation about custom development. It proves personalisation is achievable here, which makes its absence in the other five a choice, not a limitation.
Hi [Name], fintech and healthtech companies often compete for very specific talent...
Before 3.5/10→After 9.0/10
Opening lineFintech and healthtech companies often compete for very specific talent, but the bigger challenge sometimes starts after the hire: onboarding, knowledge transfer and getting people productive quickly.
Hi [Name],
Fintech and healthtech companies often compete for very specific talent, but the bigger challenge sometimes starts after the hire: onboarding, knowledge transfer and getting people productive quickly.
I'm currently exploring how product companies structure these processes as they scale.
Would be interesting to hear how you approach it at [Company]. Are your HR, recruiting, career development and operations processes covered with any AI-powered tool?
Rebuilt opening lineYour last fintech hire took three weeks to reach the point where they could touch production without a second sign-off. That gap is the expensive part, not the search.
Hi [Name],
Your last fintech or healthtech hire probably cleared the interview loop fast. The slow part came after: getting them cleared, briefed and trusted enough to touch anything real.
That gap, between start date and actually productive, is where most of the cost sits and almost nobody tracks it.
I can send you the three questions we ask HR and ops leaders to size that gap in ten minutes, no call required.
The opening sentence is a category observation, not a reader-specific consequence. "Fintech and healthtech companies often compete for very specific talent" is true of the entire industry and therefore true of no one in particular.
"I'm currently exploring how product companies structure these processes" is research framing. It tells the reader they are a data point, not the reason for the message.
"Would be interesting to hear how you approach it at [Company]."
This is the entire ask. It is answerable with one word and carries no cost to ignore. There is nothing on the table for the reader to say yes to.
What the rebuild does
The opening line names a specific, ownable moment: the reader's own last hire, and the exact friction point (production access, sign-off) rather than an industry generality.
The research framing is removed entirely. The company is not mentioned until the offer.
The close is a concrete, low-effort artifact ("three questions," "ten minutes") instead of a 20-minute calendar commitment, which is a smaller yes and a faster one.
01Opening line. Industry observation replaced with the reader's own last hire.
02Lead ownership. "I'm exploring" removed, no pivot to the company before the offer.
03Translation. "Getting people productive quickly" became "the gap almost nobody tracks."
04Proof. No invented number added, since none is verified; the offer itself (a diagnostic, not a claim) carries the weight instead.
05Repetition risk. Structure deliberately diverges from messages 02 and 03 so the same reader cannot recognise the shape twice.
06CTA. "Would be interesting to hear" became a named, low-friction artifact the reader can say yes to in one line.
07Personalisation. [Company] placeholder kept for merge, but the sentence now works even if it is never filled in.
Rebuild effort ModerateClarity after rebuild 9/10Same segment, same product. The pivot sentence is gone, the ask is smaller.
Message 02 · CEO / COO of IT service companies · Option 1
Hi {{Name}}, a quick question. How easy is it today to get a real-time view...
Before 3/10→After 9.0/10
Opening lineA quick question.
Hi {{Name}},
A quick question.
How easy is it today to get a real-time view of resource availability, staffing readiness and operational capacity across your organization?
As part of building Itelgie, an AI-powered platform for operational visibility and process automation, we're speaking with CEOs and COOs of service companies to better understand the operational challenges organizations face as they scale.
We'll be sharing a benchmarking report with key findings, trends and recommendations with all participants.
Would you be open to a short 20-minute conversation?
Rebuilt opening lineHow easy is it today to say, in one sentence, which of your delivery teams has spare capacity this week?
Hi {{Name}},
How easy is it today to say, in one sentence, which of your delivery teams has spare capacity this week?
Most COOs we talk to can answer for one team. Fewer can answer for the whole org without pulling three spreadsheets first.
We built Itelgie because we hit the same wall running our own 500-person delivery org. It is now the tool we use internally for exactly that question.
Send me your team count and I'll tell you, honestly, whether it's worth 20 minutes.
"A quick question" is a sentence that delays the actual question by one line. It signals low confidence in the question itself.
"As part of building Itelgie, an AI-powered platform for operational visibility and process automation, we're speaking with CEOs and COOs..."
This is the pivot. The reader's question in line one is dropped and replaced with a description of what Serge's company is doing. From here the email is about Itelgie, not about the reader's Monday morning.
"We'll be sharing a benchmarking report" is offered as an existing asset with no evidence it has been built. If a skeptical COO asks to see it first, there is nothing to send.
"Would you be open to a short 20-minute conversation" is the same close as message 03, sent to a different persona. A COO connected to both an HR and an ops contact at Serge's company would see the same sentence twice.
What the rebuild does
The diagnostic question is kept, since it is genuinely strong, but nothing interrupts it before the reader answers it in their own head.
"Most COOs we talk to can answer for one team" replaces the research-framing sentence with a specific, believable observation instead of a company announcement.
Itelgie is introduced as something built to solve Serge's own operational pain, which is closer to proof than a claim about the market.
The close reverses the qualification: instead of asking the reader to commit 20 minutes on faith, it offers to tell them honestly whether it's worth it, which is a smaller, safer yes.
01Opening line. "A quick question" filler cut, the question itself now opens the message.
03Translation. "Operational visibility and process automation" became "which team has spare capacity this week."
04Proof. Unbuilt benchmarking report replaced with Serge's own internal use case, which is verifiable.
05Repetition risk. Structure now diverges from message 03's HRD version instead of mirroring it.
06CTA. "Would you be open to a conversation" became an offer the reader can qualify in one line, no calendar commitment.
07Audience signal. "This is the tool we use internally" speaks to an operator, not a survey participant.
Rebuild effort LightClarity after rebuild 9/10Same question, same offer. The market-research sentence in the middle is gone.
Message 03 · HRD of IT service companies · Option 1
Hi {{Name}}, a quick question. As your organization grows, how easy is it today...
Before 3/10→After 8.5/10
Opening lineA quick question.
Hi {{Name}},
A quick question.
As your organization grows, how easy is it today to maintain visibility across workforce planning, hiring needs, employee skills and resource availability?
As part of building Itelgie, an AI-powered platform connecting HR, recruiting and business operations, we're speaking with HR leaders to better understand the challenges organizations face as they scale.
We'll be sharing a benchmarking report with key findings and practical recommendations with all participants.
Would you be open to a short 20-minute conversation?
Rebuilt opening lineWhen a project needs a fifth engineer next month, how many hours does it take your team to find out who's actually free?
Hi {{Name}},
When a project needs a fifth engineer next month, how many hours does it take your team to find out who's actually free, skilled and not already double-booked?
At most service companies we've talked to, the answer involves three tools and a Slack thread. We built Itelgie after living that exact thread ourselves.
I put together a one-page version of what we automate first. Want me to send it over, no call attached?
This message and message 02 share the identical skeleton: filler opener, diagnostic question, company pivot, unbuilt report, 20-minute close. Only the noun changes from "operational capacity" to "workforce planning."
That repetition is invisible inside a single message and expensive across a list. An HR leader and their COO, both contacted from the same sequence, would recognise the second message before finishing it.
"We'll be sharing a benchmarking report with key findings and practical recommendations with all participants."
Same unverified incentive as message 02, offered again as if it already exists.
What the rebuild does
The question is rewritten around a moment an HR director has actually lived, staffing a project on short notice, rather than an abstract "visibility" concept.
The structure is deliberately built to not mirror message 02's rebuild, even though both come from the same underlying product, so the pattern-recognition risk drops for anyone who sees both.
The close swaps the report for a one-pager, a smaller and more credible asset to promise, and removes the calendar ask entirely.
01Opening line. Filler and abstract question replaced with a lived staffing scenario.
02Lead ownership. Company pivot removed.
03Translation. "Workforce planning visibility" became "who's actually free."
04Proof. Unbuilt report replaced with a smaller, real artifact.
05Repetition risk. Rebuilt structure now differs from message 02's rebuild, not just its noun.
06CTA. 20-minute call became a one-line yes for a document.
07Audience signal. "We built it after living that exact thread" speaks to an HR operator's daily reality, not a survey.
Rebuild effort ModerateClarity after rebuild 8.5/10The fix here is structural distance from message 02, not just better copy.
Message 04 · Lead magnet · LinkedIn invite note
Hi [Name], hiring delays are expensive, but they usually don't send an invoice :)
Before 6.0/10→After 9.5/10
Opening lineHi [Name], hiring delays are expensive, but they usually don't send an invoice :)
Hi [Name], hiring delays are expensive, but they usually don't send an invoice :)
We built a simple Hiring Delay Cost Calculator to make that cost more visible. Thought it might be relevant for your team.
You can find it by the link below, or this and 11 more calculators can be also found on Itelgie official website or LI page.
Happy to connect.
LinkedIn invite notes do not support clickable links. "You can find it by the link below" points at nothing the reader can tap. This is a structural failure, not a wording one, and it caps this message's score no matter how the copy reads.
Rebuilt opening lineHi [Name], hiring delays are expensive, they just don't send an invoice.
Hi [Name], hiring delays are expensive, they just don't send an invoice.
We built a Hiring Delay Cost Calculator that puts a number on it. Two minutes, your own figures, nothing to install.
Accept and I'll send the calculator straight over, along with the eleven others on the same theme.
The opening line is genuinely good: consequence-first, specific, has a wink of tone without trying too hard. This is the strongest first sentence in the set.
"Thought it might be relevant for your team" is a hedge. It downgrades a sharp opener into a maybe, right before the ask.
"You can find it by the link below" cannot work in a LinkedIn connection note. There is no link field. The reader has nowhere to go even if they want to.
"Happy to connect."
This is a banned phrase in the house style and, more importantly, guest language: it states Serge's hope, not what the reader gets by accepting.
What the rebuild does
The opener is kept almost verbatim, since it earns its place.
The hedge is removed. "Puts a number on it" replaces "thought it might be relevant."
The broken link promise is replaced with a clear, deliverable next step: accept, then the calculator arrives by message, which is the only channel that actually works at this stage of the funnel.
The close names the reader's own outcome, "my number," instead of Serge's wish to connect.
01Opening line. Kept, it was already the best line in the set.
02Hedge removal. "Might be relevant" cut.
03Translation. "Cost more visible" became "puts a number on it," concrete over abstract.
04Structural fix. Dead link promise replaced with a deliverable-after-accept flow that actually works in a connection note.
05Proof. "Two minutes, your own figures" sets honest expectations instead of overselling.
06CTA. "Happy to connect" became "Show me my number," the reader's decision, phrased as they'd phrase it.
07Audience signal. Asset-led message needs no persona pivot, which is exactly why it already outscored the other four.
Rebuild effort LightClarity after rebuild 9.5/10Best message in the set before the rebuild. The fix is removing a broken promise, not rewriting the idea.
Message 05 · Warm reconnect · past SoftTeco custom-dev leads
Hi [Name], when we last spoke, the conversation was mostly about custom development.
Before 6.0/10→After 9.0/10
Opening lineHi [Name], when we last spoke, the conversation was mostly about custom development.
Hi [Name], when we last spoke, the conversation was mostly about custom development.
Since then, we have taken one of the problems we regularly saw in growing companies - fragmented HR and operational processes - and built an actual product around it.
It is called Itelgie, and it brings recruiting, employee management, career growth, projects, approvals and analytics into one system.
You seemed like exactly the type of company it was created for, so I wanted to reconnect. Curious to see it?
Rebuilt opening lineHi [Name], last time we spoke it was about custom dev. Something's changed since.
Hi [Name], last time we spoke it was about custom dev. Something's changed since.
We kept running into the same problem across growing clients: HR, recruiting, approvals and reporting living in five different places. So we built a product for it instead of quoting another custom build.
It's called Itelgie. Recruiting, employee management, career growth, projects, approvals and analytics, one system.
You brought this up yourself the last time we talked. Want the two-minute version?
This is the only message in the set with real personalisation, a genuine reference to a past conversation, and it shows in the score. The bones are right.
"You seemed like exactly the type of company it was created for..."
"Exactly" is a banned word in house style and also does real work here: it overclaims certainty about a company Serge is inferring, not one that has confirmed fit.
"Curious to see it?" is soft. It asks about the reader's curiosity rather than offering a specific, small next step, so a busy reader can let it sit unanswered without feeling they said no to anything concrete.
What the rebuild does
The personalisation is kept and sharpened: "you brought this up yourself" reframes it as the reader's own words coming back, not Serge's assumption about fit.
"Exactly the type of company" is replaced with a claim Serge can actually stand behind.
The close becomes a named, small artifact, "the two-minute version," instead of an open-ended question about curiosity.
01Opening line. Kept, personalisation is the strength of this message.
02Overclaim removed. "Exactly the type of company" replaced with the reader's own past words.
03Translation. "Fragmented HR and operational processes" kept, since it was already plain language.
04Proof. The past conversation itself is the proof point, correctly used here unlike anywhere else in the set.
05CTA. "Curious to see it?" became a named, sizeable-in-advance artifact.
06Audience signal. Already the strongest in the set; the rebuild protects it rather than replacing it.
07Word choice. Banned word "exactly" removed per house style.
Rebuild effort LightClarity after rebuild 9/10Second-best message in the set. Proof this segment can be personalised, the other five should copy this one, not the reverse.
Message 06 · Screenshot-led message
Hi [Name], instead of sending you another long description of what Itelgie does...
Before 7.0/10→After 9.5/10
Opening lineHi [Name], instead of sending you another long description of what Itelgie does, I'll leave this screenshot here :)
Hi [Name], instead of sending you another long description of what Itelgie does, I'll leave this screenshot here :)
These are just a few real workflows we automate — hiring, onboarding, employee management, time tracking and internal communication.
The point is not to add one more tool. It is to reduce the number of tools, spreadsheets and manual updates your team already has.
Anything here looks painfully familiar?
Rebuilt opening lineHi [Name], instead of another paragraph about what Itelgie does, here's the actual screen.
Hi [Name], instead of another paragraph about what Itelgie does, here's the actual screen.
These are five real workflows it runs today: hiring, onboarding, employee management, time tracking, internal communication.
The goal isn't one more tool. It's fewer tools, fewer spreadsheets, fewer manual updates than your team already has.
Anything here look painfully familiar? Tell me which one and I'll show you exactly that part running.
This is the second-strongest message in the set, and for the right reasons: it shows instead of describes, it's honest about being tired of long descriptions, and it closes on a question that names the reader's actual pain instead of asking for their time.
"a few real workflows we automate — hiring, onboarding, employee management, time tracking and internal communication"
The one soft spot: the closing question has no defined next step. If the reader says "yes, the time tracking one," there's nothing telling them what happens after they say it.
The em dash in the workflow list is a small house-style miss, not a conversion issue.
What the rebuild does
Keeps the structure almost entirely intact, since it was already close to right.
Adds a named next step after the closing question, so a "yes" has somewhere concrete to go instead of dead-ending in a reply that Serge then has to manually follow up.
Replaces the em dash with a comma-separated list per house style, with no change to meaning.
01Opening line. Kept, tightened by one word.
02Lead ownership. Already reader-first, no change needed.
03Translation. Already concrete ("hiring, onboarding, time tracking"), kept as is.
04Proof. The screenshot itself functions as proof, the strongest proof mechanism anywhere in the set.
05CTA. Open question given a defined next step so a "yes" doesn't stall.
06Audience signal. "Painfully familiar" already speaks to a specific operational reader, unchanged.
07Style. Em dash removed per house punctuation rule.
Rebuild effort LightClarity after rebuild 9.5/10Strongest message in the set. The fix is giving the "yes" a destination, not rewriting the idea.
How each message was scored
Every message ran through the same seven checks, adapted from the Decision Friction Model for cold outreach rather than lifecycle email. Each check is scored out of ten, and the average becomes the message's score. The order matters: a reader who stops after the pivot sentence never reaches the close, so the company-first pivot is the single most expensive failure in this set, not the weakest CTA.
01Opening line: filler and category talk, or the reader's own moment"A quick question" and "Fintech companies often compete for talent" describe a category. A reader-owned scenario, a specific hire, a specific project, a specific past conversation, is what earns a reply.
02Lead ownership: does the message stay on the reader past line twoThree of six messages pivot to "as part of building Itelgie, we're speaking with X" by line three. The reader's problem is dropped the moment the product enters.
03Feature-to-outcome translation: platform description or named outcome"An AI-powered platform for operational visibility and process automation" describes the product. "Which team has spare capacity this week" describes what the reader gets.
04Proof: verified detail or unbuilt incentiveA benchmarking report offered as fact, with no evidence it exists, is a claim, not proof. Serge's own internal use of Itelgie, or a real past conversation with the reader, is proof.
05Structural repetition: distinguishable message, or the same skeletonMessages 02 and 03 are close to identical in shape. A reader who receives both, through two different contacts at the same company, will notice.
06CTA: ownership language or guest language"Would you be open to a conversation" and "Happy to connect" describe Serge's hope. "Show me my number" and "Send me the two-minute version" describe the reader's decision.
07Personalisation: written for this reader, or written for the segmentOnly the reconnect message references something the specific recipient actually said. Every other message would read identically to any name dropped into the placeholder.
Scoring is deliberately harsh on structure and generous on tone. Messages 04 and 06 are both well-written and well-scored, they broke the pattern instead of following it. Messages 02 and 03 are also well-written and scored lowest, because the pattern, not the prose, is what costs the reply.
What the numbers actually mean
Every score on this page is out of ten. Across 67 previous Strategic Flow teardowns the average opening score was 3.4 and the average after rebuild was 9.0, so the range below is calibrated against real sends rather than a theoretical ideal.
1 – 3Structurally brokenNo reason to reply and no clear next step even if the reader wants to. Nothing in this set is this bad.
3.5 – 5.5Competent but leakingProfessionally written, nothing embarrassing, but the pivot to the product arrives too early and the close asks for too much too soon. Messages 01, 02 and 03 sit here.
6 – 7.5Working, with one weak linkStrong opening, real personalisation or a real artifact, let down by a soft close or a structural gap. Messages 04, 05 and 06 sit here.
8 – 10Reader-orderedOpens on the reader's own moment, stays there past the pivot point, and closes on a small, named yes.
Rebuild effort is noted on each message separately. Two were rated Light, since the ideas were already right and only the pivot or the close needed cutting. Three were rated Moderate, since the underlying skeleton itself, not just a line, needed to change.
The finding that matters most
Individually, these six read as reasonable outreach from a 500-person engineering company testing a new product. The writing is clear throughout and two of the six, the lead magnet invite and the screenshot message, are close to ready as they stand.
The cost sits in three structural habits, visible only across the set:
The same skeleton was reused across three ICP segments. FinTech/Healthtech, CEO/COO and HRD messages open with a diagnostic question, pivot to "as part of building Itelgie, we're speaking with X," and close on a 20-minute ask. That skeleton is efficient to write and expensive to send, because any reader connected to more than one contact at Serge's company will recognise it before the second message finishes.
The pivot happens one sentence too early. In three of six messages, the reader's problem is stated in line one and abandoned by line three, replaced with a description of what Serge's company is building. The strongest lines in the whole set, the last hire that took three weeks, the fifth engineer nobody can locate, the screenshot instead of another paragraph, all stay on the reader's side the entire time.
An unbuilt incentive is offered as an existing asset. "We'll share a benchmarking report with all participants" appears three times with nothing in the thread confirming it has been assembled. This is the one place in the set that risks crossing from optimistic copy into a claim Serge can't back up if a careful CEO asks to see it before replying.
Personalisation exists exactly once, and it's the second-highest-scoring message. The reconnect email references a real past conversation. Nothing structurally stops the other five from doing the same, even at a lighter touch, a shared connection, a company update, a role change. The absence in five of six is a resourcing choice, not a limitation of cold outreach.
The rebuilds above change no product claims and no pricing. Same Itelgie, same ICPs, same list. Only the order in which each reader meets the company, and how small the first yes is.
Strategic Flow · Decision Friction Model
Seven-point diagnostic · 67 teardowns published strategicflow.tech