The Fable 5.1 Client Acquisition Playbook: 6 Stages, 15 Skills, One Audit That Closes

Most people read a model release as a capability update. For anyone selling services it is a pricing and packaging event.

Three things in Fable 5.1 change what you can put on an invoice. Everything else in this playbook is downstream of them.

Change 1: multi-step workflows became sellable

The business workflows benchmark went from 17.1% to 31.4%. That measures whether a model can reliably automate a real pipeline end to end.

At 17% you cannot retainer that. You build it, it breaks on the fourth run, and the margin goes on fixing it. At 31% it fails often enough to need monitoring and rarely enough to charge for.

Newly sellable, in order of how easy they are to close:

1. Proposal generation from a form. Every agency has this pain and nobody has solved it for them.
2. Weekly client reporting that assembles itself from real data.
3. Content repurposing pipelines, one long asset into a month of channel content.
4. Lead routing with real enrichment and CRM writeback.
5. Inbox triage and reply drafting across accounts.

Sell these as installed systems with a monitoring retainer, not one-off builds. The build is the wedge. The monitoring is the business.

Change 2: three verticals reopened

Fable 5 used to fall back to a smaller model mid-task on anything near biology, medicine or security. Unpredictably, sometimes on a completely benign question.

Fable 5.1 flags benign requests 60% less often, and on biology and medical questions the fallback rate is down 85%.

That is a client acquisition fact rather than a technical one. Health, medical, dental, insurance, finance and security clients were painful to serve, so most agencies quietly deprioritised them. Those are also the verticals with the biggest budgets and the fewest competitors willing to do AI work in them.

The play: go after the vertical your competitors abandoned. Lead with the tooling problem being solved and the compliance discipline being yours.

One caveat to build into the offer rather than hide: human review before publication stays non-negotiable in regulated sectors. Price it as a QA line item and it becomes a selling point instead of a cost.

Change 3: cost per task, not cost per hour

Roughly 25% cheaper on standard workloads, up to 45% on agentic ones, cache reads down around 75%, and roughly 2.5x the output per dollar against Fable 5 on Anthropic's own curve.

The pricing implication most people miss: your delivery cost per client dropped while your prices did not. That gap is either margin or a competitive weapon, and you have to pick.

Three ways to spend it:

• Keep it as margin. Simplest, and defensible if your delivery quality is the differentiator.
• Undercut on the wedge only. Free audit, cheap pilot, normal retainer. Lowers the cost of getting in the door without devaluing the ongoing work.
• Widen the offer. Include the reporting workflow at no extra charge because it now costs you almost nothing to run. Increases perceived value without touching price.

The one to avoid is cutting retainer prices. The saving is in tokens, and a client who gets a discount for a model release will expect another one at the next release.

The offer, written out

Run this first. Everything downstream reads the output.

mkdir -p .claude/skills/offer-architect

---
name: offer-architect
description: Define a services offer that is specific enough to sell.
  Use before any outreach or lead magnet work.
---

Take ownership of this from start to finish.

GOAL: one offer document I could send to a stranger who would
understand exactly what they get and what it costs.

Ask me, in ONE message, for anything missing:
1. Which of the five newly sellable workflows I want to lead with
2. Which vertical, exactly. Not "B2B". "Multi-site dental practices
   in the UK" is a vertical.
3. What I have already delivered that I can reference, with numbers
4. What I refuse to do
5. My delivery cost per client per month, roughly

Then output:

THE OFFER
- One sentence a stranger understands
- What they receive, as objects. Files, dashboards, workflows.
- What happens in week 1, week 2, week 4
- What is explicitly not included
- The price, and what it does not cover

THE WEDGE
- The free thing that proves competence before they pay
- Why it is genuinely useful even if they never buy

THE DISQUALIFIER
- Who this is not for, stated openly
- Naming this raises close rate. An offer for everyone converts
  nobody.

HARD RULES
- Never invent a result, client or figure. Only what I gave you.
- No "bespoke" or "tailored". Those words mean the offer is not
  defined.
- The price appears. An offer with no number is a conversation, not
  an offer.

Before finishing, review this as a skeptical buyer in that vertical.
Find the five weakest points. Fix them. Then verify a stranger could
act on it.

Write to acquisition/offer.md.

The disqualifier is the section people cut and the one that lifts close rate most. A prospect who reads "not for you if you have under three locations" and keeps reading has qualified themselves.

Four skills. The list is not the hard part. Finding the ones with a reason to move this quarter is.

Run the filter on low effort. Run the signal work on high. That split is where the cost saving actually lands.

2. target-list-builder

mkdir -p .claude/skills/target-list-builder

---
name: target-list-builder
description: Build a raw target list for a services offer. Use after
  the offer exists.
---

Read acquisition/offer.md for the vertical and the disqualifier.

Source in priority order, highest intent first:
1. Commenters and likers on posts about the problem your offer solves,
   including competitor posts
2. Attendees of relevant industry events
3. Members of relevant industry groups
4. Search matching the vertical, by job title and company size
5. Job boards, for companies hiring the role your offer replaces

Source 5 is underused for services. A company posting for a marketing
ops hire has budget and an unmet need at the same time.

Capture: name, headline, company, size, location, LinkedIn URL,
source.

Write to acquisition/prospects.csv with a linkedin_url column. That
header name is required by the import later.

Enrich nothing. Look up no emails. Most of this gets discarded next.

Source 1 is the one to run first. Prosp imports commenters and likers straight from any post URL, including future reactions as they arrive, so a competitor's post that keeps performing keeps feeding your list without you touching it.

3. buying-signal-finder

Services buying signals are different from product ones. You are looking for evidence someone has the problem and has decided to spend on it.

mkdir -p .claude/skills/buying-signal-finder

---
name: buying-signal-finder
description: Find a dated, verifiable reason each target would buy
  services this quarter.
---

One signal per target. Ranked by how close it is to a budget
decision.

TIER 1, budget is already moving
- Hiring for the role your offer replaces or supports
- Just raised, last 90 days
- New marketing or ops leader in post, last 60 days
- Publicly complained about the exact problem, last 30 days
- Running paid ads badly, visibly and currently

TIER 2, the problem is visible
- Site or content shows the specific gap your offer fixes
- Reviews or comments mention the problem
- Expanding into a new market or location
- Competitor of an existing client of yours

TIER 3, fit only
- Matches the vertical, no observable trigger

Every signal verifiable and dated. Record the date and URL, because
you will reference it in the first message and they will check.

No signal means write NONE. Never stretch "is a dentist" into a
trigger.

A new leader in post is the strongest services signal there is. They
have ninety days to show a change and no loyalty to the incumbent.

4. fit-filter

Low effort, cheap model, runs across the whole list. This is the highest-volume call in the playbook and has the least reasoning in it.

mkdir -p .claude/skills/fit-filter

---
name: fit-filter
description: Filter a raw target list against the offer before any
  enrichment. Run on Haiku at low effort.
---

Read acquisition/offer.md, especially the disqualifier.

Per row return exactly one of KEEP, DROP, UNSURE.

KEEP    Matches the vertical, the size band and the buyer role
DROP    Fails the disqualifier, or is a competitor, or is another
        agency when you sell to end clients, or is too small to
        afford the stated price
UNSURE  Headline is ambiguous. Do not guess.

One line of reasoning per row, 12 words maximum.

KEEP and UNSURE to acquisition/filtered.csv.
DROP to acquisition/dropped.csv so the filter can be audited.

Report the drop rate. Below 40% usually means the search was too
narrow rather than the list being unusually good.

The price check in DROP matters. A company that cannot afford the
number in offer.md is not a lead, it is a wasted send.

5. list-grader

The gate. Nothing goes out against a list below B.

mkdir -p .claude/skills/list-grader

---
name: list-grader
description: Grade a target list A to F before it enters outreach.
---

Grade acquisition/filtered.csv across eight dimensions, 1 to 10:

1 Vertical fit
2 Buyer role accuracy: decision makers or influencers
3 Affordability against the price in offer.md
4 Data completeness
5 Deduplication, within the list and against past outreach
6 Recency of activity
7 Reachability, how many are already 1st degree
8 Signal strength, how many carry a Tier 1 or 2 signal

Average into a grade.
A 9.0+ send today   B 7.5+ send   C 6.0+ fix the two lowest first
D 4.5+ rebuild the list   F below rebuild the offer, not the list

Below B, stop. Name the two lowest dimensions and the fix for each.

An F grade usually means the offer is too broad, not that the list is
bad. Go back to stage 1.

Dimension 8 is the one that predicts reply rate. A B-grade list where most rows carry a Tier 1 signal outperforms an A-grade list with none.

Why filter before enrich

The expensive mistake is enriching a thousand rows to discover six hundred were never going to buy.

Order matters: build raw, filter cheap, grade, then enrich only what survived. Enrichment is per-lookup pricing and the filter costs almost nothing at low effort.

Dimension 7, reachability, also changes the campaign shape rather than the list quality. Mostly 1st degree can be messaged directly. Mostly cold needs connection requests first, and that is a different budget: 20 requests a day against roughly 100 messages a week. Prosp handles both in one sequence with a lead is 1st level condition, so you do not need two campaigns.

The highest-converting services lead magnet is not a guide. It is a piece of work already done on their business, unasked.

This is where the cost reduction pays for itself. An audit that used to be too expensive to give away for free now costs you a fraction of a retainer hour to produce.

Why the audit beats every other wedge

A guide asks them to do the work. An audit shows them the work already done, on their own account, with their own numbers in it.

Three things it does that nothing else does:

1. It proves competence without a case study. Especially useful if you are new, or entering a vertical where you have no logos yet.
2. It creates the problem before you pitch the solution. They cannot unsee a list of things wrong with their own site.
3. It qualifies hard. Someone who reads a nine-page audit of their own business and replies is a different prospect from someone who downloads a PDF.

The economics only work at Fable 5.1 pricing. At the old cost per task, giving away a real audit to fifty prospects was a marketing budget. Now it is closer to a rounding error, and you still charge full price for the fix.

6. free-audit-generator

Run this with the finish-the-job wrapper. It is a multi-step task and it is exactly what 5.1 is built for.

mkdir -p .claude/skills/free-audit-generator

---
name: free-audit-generator
description: Produce a free audit of a prospect's business as an
  outreach wedge. Use on Tier 1 and Tier 2 signal targets only.
---

Take ownership of this from start to finish.

GOAL: a document a prospect would find useful even if they never
reply, containing findings specific to their business.

INPUT: their website, their LinkedIn, the signal found in stage 2,
and acquisition/offer.md for what you are qualified to comment on.

STRUCTURE, six sections

1. WHAT YOU ARE DOING WELL
   Two genuine things. Specific, not flattery. This is not padding,
   it is proof you actually looked.

2. THE THING COSTING YOU MOST
   One finding, the biggest. What it is, why it costs them, and
   roughly what it costs. Only quantify if you can defend the
   arithmetic from something observable.

3. FOUR MORE FINDINGS
   Each: the issue, the evidence with a URL or screenshot reference,
   the fix, and the effort. Ranked by impact.

4. WHAT I WOULD DO FIRST
   One action. Not a list. Something they could start on Monday
   without hiring anyone.

5. WHAT THIS AUDIT DOES NOT COVER
   State the limits honestly. What you would need access to in order
   to say more. This section is what makes the rest credible.

6. IF YOU WANT THE REST
   Two lines. What the paid engagement covers and the price.

HARD RULES
- Every finding needs observable evidence. Never assert a metric you
  cannot see.
- Never invent a benchmark or an industry average.
- Section 4 must be something they can do WITHOUT you. That is what
  makes it a gift rather than a sales document.
- No em dashes. Under 900 words total.

Before finishing, review it as a skeptical buyer. Find the five
weakest findings and either strengthen them with better evidence or
cut them. Four strong findings beat eight thin ones.

Verify every claim traces to something you actually observed.

Section 5 is the counterintuitive one. Naming what you could not assess is what makes a prospect trust the parts you did.

7. audit-to-offer-bridge

The audit gets read. Then nothing happens, unless the bridge exists.

mkdir -p .claude/skills/audit-to-offer-bridge

---
name: audit-to-offer-bridge
description: Convert a delivered audit into a booked call. Use 2 days
  after the audit lands.
---

Read the audit and acquisition/offer.md.

Write one message, under 300 characters.

IT MUST
- Reference the specific finding from section 2, by name
- Ask one question about whether that finding is already known to
  them internally
- Not pitch, not mention the price, not ask for a call yet

The question is the whole message. "Was the [specific thing] already
on your radar, or is that new?" gets replies because it is answerable
in four words and it is genuinely interesting to them.

IF THEY REPLY THAT IT IS KNOWN
  They have a stalled internal project. Ask what is blocking it. The
  blocker is your entry point.

IF THEY REPLY THAT IT IS NEW
  Offer to walk through the other findings. 15 minutes, three
  questions, no deck. Then the calendar link.

IF NO REPLY AFTER 4 DAYS
  One more message. A second finding they have not seen, no ask.
  Then stop.

Never send a third. Two unanswered messages after a free audit means
the timing is wrong, not that the message was.

8. audit-at-scale

One audit is a nice gesture. Fifty is a pipeline. This is the workflow that makes it repeatable.

mkdir -p .claude/skills/audit-at-scale

---
name: audit-at-scale
description: Batch-produce audits across a graded target list.
---

Read acquisition/filtered.csv. Process only rows with a Tier 1 or
Tier 2 signal. Tier 3 does not get an audit, it gets a normal
sequence.

Per row:
1. Run free-audit-generator
2. Save to acquisition/audits/[company]-[YYYYMMDD].md
3. Log to acquisition/audit-log.csv: company, contact, LinkedIn URL,
   date, the section 2 finding in one line, and status = READY

Before batching, run ONE audit and show it to me. Do not produce
fifty against an unvalidated template.

AFTER I APPROVE
  Batch the rest. Report how many produced, how many were skipped
  for insufficient observable evidence, and why.

A skipped audit is a correct outcome. Some prospects have nothing
visible to audit, and a thin audit is worse than none.

RUN AT MEDIUM EFFORT per audit. High effort on a batch of fifty is
where the cost saving disappears.

The one-then-approve step is the difference between a batch that converts and fifty documents you have to bin. The template only becomes reliable after you have read one properly.

The delivery mechanism

Fifty audits sitting in a folder are worth nothing. The bottleneck moves to delivery the moment the production problem is solved.

The route that works: a post about the problem the audit uncovers, with a comment CTA, then delivery by DM to everyone who engages. That way the audits go to people who have already raised a hand rather than to a cold list.

Prosp imports the commenters and likers straight from the post URL, then runs the sequence: a lead is 1st level condition, the message with the audit link on the YES branch, a connection request then the message on the NO branch, and a Reply Comment node so every public comment gets a public reply too.

Connection requests auto-detect acceptance, checking every 24 hours for two weeks, so no wait card is needed after that step.

For the outbound version, import filtered.csv directly. It needs headers in row one and a linkedin_url column, and the audit link can travel as a custom variable so each message carries that prospect's own document.

Four skills. Two sequences: one for prospects who got an audit, one for those who did not.

The audit branch converts several times better, so send capacity goes there first and the Tier 3 branch gets whatever is left.

9. opener-writer

mkdir -p .claude/skills/opener-writer

---
name: opener-writer
description: Write first-touch messages for a services offer, from a
  buying signal.
---

Read the signal, and acquisition/offer.md.

RULES
- Under 300 characters. Longer reads as a template.
- Line one references the signal specifically. Not "saw you're
  hiring" but the actual role and what it implies.
- One ask, answerable in under ten words.
- Never mention your price, your process or your agency name in
  message one.
- No compliments. No "hope this finds you well".
- No em dashes.

STRUCTURE
Line 1  the signal, dated and specific
Line 2  the inference you drew from it
Line 3  the question

WRITE ONE PER TIER
Tier 1  lead with the signal, it is recent and they know it
Tier 2  lead with the observable problem, since there is no event
Tier 3  lead with the vertical-level pattern, no personalisation
        claim you cannot back

BANNED, these are burnt in services outreach:
"I came across your profile"
"I help [vertical] with [service]"
"Are you open to a quick chat"
"I noticed you're in [industry]"

The second banned line is the one everyone uses. "I help dentists with marketing" tells them you have a list, not that you looked at them.

10. sequence-builder

mkdir -p .claude/skills/sequence-builder

---
name: sequence-builder
description: Build the outreach sequence spec for both branches.
---

Output a node list, not prose. Per node: action, delay before it,
exact copy, exit condition.

BRANCH A, AUDIT SENT
  1  Message: the audit link, one line of context
  2  Wait 2 days
  3  Message: audit-to-offer-bridge, the section 2 finding question
  4  Wait 4 days
  5  Message: a second finding, no ask
  Stop. Two unanswered after an audit means timing, not message.

BRANCH B, NO AUDIT (Tier 3)
  1  Message: the vertical-level opener
  2  Wait 3 days
  3  Message: a different angle on the same problem, no link
  4  Wait 5 days
  5  Message: the close-out. "Assuming this isn't a priority right
     now, which is fine. Want me to check back in Q[X]?"
  Stop.

RULES
- Reply stops the sequence for that lead, always.
- A connection request auto-detects acceptance, so no wait node
  after one.
- Every touch a different angle. Never "just following up".
- Any voice note pairs with a written message.
- Branch on connection state first: 1st degree messages directly,
  cold gets a connection request then the message.

The branch B close-out gets more replies than the two before it
combined, because permission to say no removes the pressure that was
stopping them answering.

This maps one to one onto the Prosp builder. Actions: connection request, message, voice note, InMail, message to open profile, comment on last post, reply comment, like last post, visit profile, wait X days, add tag. Conditions: has LinkedIn URL, lead is 1st level, opened message, lead is open profile, check data in column.

Use add tag to mark which branch each lead is on, so reporting can compare audit conversion against no-audit conversion. That single number decides whether the audit is worth producing at scale.

11. personalisation-prompt-writer

mkdir -p .claude/skills/personalisation-prompt-writer

---
name: personalisation-prompt-writer
description: Convert sequence copy into prompts that personalise per
  lead at send time.
---

Syntax, keep the two distinct:
  {{double curly}}   the tool's own profile data
  [square brackets]  instructions to the AI

Variables: first name, last name, headline, biography, job title,
company, last post. Plus any custom variable from the CSV, so the
signal and the audit link are both usable.

RULES
- Only the opening line varies. Keep the ask identical so branch
  performance is comparable.
- State length in characters explicitly.
- Say what to do when a variable is empty. Never allow "Hi ,".
- Ban the burnt openers inside the prompt itself.
- One instruction per bracket.

SHAPE
  Hi {{first name}},
  [One sentence referencing {{signal}}. Be specific about the actual
  detail, not the category. Under 90 characters. If {{signal}} is
  empty use {{headline}}. Never open with "I came across" or
  "I help".]
  [Then this line unchanged:] Worth a look at what I found?
  [Then:] {{audit_link}}

Preview against real contacts before sending. That catches a prompt
that reads fine on lead one and breaks on lead five.

12. reply-router

mkdir -p .claude/skills/reply-router

---
name: reply-router
description: Classify inbound replies to services outreach and route
  each one.
---

Exactly one category:

AUDIT_ENGAGED    Asked about a finding, disputed one, or asked for
                 more. The strongest signal in the system.
INTERESTED       Wants to know about the offer
TIMING           Interested, not now. Capture the date named.
OBJECTION        Price, fit, incumbent agency, internal capacity
WRONG_PERSON     Not their remit. Ask who owns it.
NOT_INTERESTED   Clear no. Thank them, stop, never pitch again.
ESCALATE         Anything below 8 confidence

Per reply: category, confidence 1 to 10, the sentence that decided
it, and the recommended next action.

AUDIT_ENGAGED goes straight to a call offer. Someone arguing with
your findings is more likely to buy than someone praising them,
because they have engaged with the substance.

Never auto-send to NOT_INTERESTED. A human closes that loop.

Do not guess at tone. Sarcasm, brevity and politeness all read the
same in text.

The AUDIT_ENGAGED category is the one worth adding to a normal reply classifier. A prospect disputing a finding has read it carefully, which is a stronger buying signal than agreement.

The send budget, and why it caps your pipeline

You can produce fifty audits in an afternoon. You cannot send fifty connection requests in one.

Recommended limits are 20 connection requests a day or 100 a week, and roughly 100 messages a week. They are account-wide, not per campaign. Three campaigns set to 20 each means 60 attempts a day and a restricted account.

So the real constraint on client acquisition is not lead generation, it is send capacity. Which means the ordering work in stage 2 decides your quarter. A list worked in the wrong order wastes weeks of budget on people who were never going to move.

Practical split for one account: 12 requests a day to audit-branch prospects, 8 to Tier 3. Set each campaign individually rather than relying on the builder defaults, which are higher than the recommended limits.

Three skills for the last stretch. The call, the proposal, and the diagnostic that tells you which stage is actually broken.

The proposal is where most services deals die, because it arrives four days later and reads like a template.

13. call-prep

mkdir -p .claude/skills/call-prep

---
name: call-prep
description: Prepare a discovery call brief. Run the morning of the
  call, not the night before.
---

Read the audit if one was sent, the signal, and the reply thread.

Output one page:

WHAT WE KNOW
- The signal, dated
- The section 2 finding from the audit, if sent
- What they said in the thread, quoted

WHAT WE DO NOT KNOW, ranked
The five gaps that would change the recommendation. Ranked by how
much they change it. These are the questions.

THE FIVE QUESTIONS
One per gap. Open, short, answerable without preparation.
Never ask something the audit already answered. That signals you did
not read your own document.

WHO ELSE IS INVOLVED
What to ask to find the other decision makers without making it
obvious you are asking.

LIKELY OBJECTIONS
Two, based on their size and sector. With the honest answer to each,
not a rebuttal.

DISQUALIFIERS
What would make this a bad client. If you hear these, say so on the
call rather than sending a proposal you hope they decline.

Under one page. A brief you cannot read in three minutes will not be
read.

The disqualifiers section is the one that protects your year. A bad client costs more than an empty slot.

14. proposal-writer

Run with the finish-the-job wrapper, same day as the call while the notes are warm.

mkdir -p .claude/skills/proposal-writer

---
name: proposal-writer
description: Write a services proposal from call notes. Same day as
  the call, always.
---

Take ownership of this from start to finish.

GOAL: a proposal they could sign without another call.

ASK ME, in one message, for anything missing:
1. The problem in their words. Quote them literally.
2. What they have already tried, and why it failed
3. Deliverables agreed on the call
4. Price to quote
5. Budget signal: named a figure / gave a range / dodged it / no
   budget yet
6. Anything that would kill this deal

FOUR SECTIONS, in order

1. DIAGNOSIS
   Reflect their problem back in their own phrasing. Not a
   paraphrase. Then the underlying cause, using why the previous
   attempt failed. Then what it costs to leave it.

2. APPROACH
   Phased. Per phase: what we do, what they receive as an object,
   roughly when. Phase one must address why the last attempt failed.
   Include one line naming what we do NOT do.

3. SCOPE AND INVESTMENT
   Branch on the budget signal:
   named a figure -> ONE option at that price. Three tiers signals
                     you were not listening.
   gave a range   -> TWO options, the quote and a 70% reduced scope.
                     Name what comes out.
   dodged it      -> THREE tiers at roughly 70%, 100%, 140%. The
                     tiers do the qualifying.
   no budget yet  -> ONE option plus a paid discovery at ~15%.
   State what the price excludes, payment terms, notice period.

4. WHAT COULD GO WRONG
   Two or three real risks, starting with the deal-killer they named.
   Each: the risk, then what we do about it. Real risks only.
   "This takes time" is a disclaimer, not a risk.

HARD RULES
- Never invent a number, client name or result.
- Timeline in week ranges, never dates. Dates go stale if it is not
  signed that week.
- Never discount unprompted.
- No em dashes.

Before finishing, review as a skeptical buyer. Find the five weakest
points. Fix them. Verify they could sign it without another call.

Section 4 is the one people cut because it feels like handing over ammunition. It does the opposite. Naming the objection first takes it off the table before it gets raised in a meeting you are not in.

15. acquisition-diagnostic

Weekly. Works the chain in order and stops at the first failure, so you fix the actual problem rather than the visible one.

mkdir -p .claude/skills/acquisition-diagnostic

---
name: acquisition-diagnostic
description: Diagnose which stage of client acquisition is broken.
  Run weekly.
---

Pull: requests sent, accepted, audits delivered, replies, calls
booked, proposals sent, deals closed. Split by branch A and B.

Work in order. STOP at the first failure. Do not report six stages
of numbers when the first one is broken.

1 ACCEPTANCE BELOW 15%
  The connection note, or the list is off-vertical. Not the
  sequence, not the audit. Do not touch those.

2 ACCEPTANCE FINE, REPLIES BELOW 10%
  The list is fine. The first message is the problem. Check: is the
  signal specific, one ask, under 300 characters.

3 REPLIES FINE, CALLS LOW
  The outreach works. The ask is mistimed, or the audit is not
  landing. Check whether AUDIT_ENGAGED replies are converting.

4 CALLS FINE, PROPOSALS NOT CONVERTING
  Not an acquisition problem. Either the price is wrong for the
  vertical or the offer does not match what they came for.

5 PROPOSALS CONVERTING, VOLUME LOW
  Nothing is broken. You are capacity-limited on sends. Check the
  daily split.

ALSO REPORT
- Branch A versus branch B conversion. If the audit branch is not
  meaningfully better, stop producing audits.
- Which signal tier produced the closed deals

ONE change per week. Three at once and a lift tells you nothing.

The branch comparison is the number that decides your process. If audits convert the same as no audits, they are a cost with no return and the honest answer is to stop.

The effort routing

Effort	Skills	Why
Low	fit-filter, reply-router	Classification across volume. One right answer per row.
Medium	opener-writer, personalisation-prompt-writer, audit-at-scale, call-prep, audit-to-offer-bridge	Known artefact from known inputs
High	offer-architect, buying-signal-finder, free-audit-generator (first run), sequence-builder, proposal-writer, acquisition-diagnostic	Judgement, planning, diagnosis

Run the first audit at high effort so the template is right, then batch the rest at medium. High effort across fifty audits is where the cost advantage disappears.

The expensive mistake is running proposal-writer low. A proposal is the highest-stakes document in the whole playbook and the tokens saved are trivial against a lost deal.

The whole thing as one run

Take ownership of this entire task from start to finish.

GOAL: a launch-ready client acquisition system, with every output in
acquisition/.

Run in order: offer-architect, target-list-builder,
buying-signal-finder, fit-filter, list-grader, free-audit-generator
(one only), opener-writer, sequence-builder,
personalisation-prompt-writer.

Stop and wait for me at three points only:
- After offer-architect, before any list building
- After the single sample audit, before batching
- After personalisation-prompt-writer, before anything is sent

Between those, keep working. Make intermediate decisions and note
them rather than checking in.

Before finishing: review the whole set as a skeptical buyer in the
target vertical. Find the five biggest weaknesses. Fix the important
ones. Then verify the result accomplishes the original goal.

HARD RULES
- Never invent a number, client name or result.
- Grade the list before it leaves. Below B, rebuild rather than
  proceed.
- If any stage cannot complete honestly, stop and name what is
  missing.

End with the three things I need to review, the daily send split, and
every assumption you made.

Three things to remember

1. The audit is the offer's proof, not its advertisement. Section 4 has to be something they can act on without you, or it reads as a sales document and converts like one.
2. Send capacity caps your pipeline, not lead generation. 20 connection requests a day is the real constraint, which makes the ordering work in stage 2 the highest-leverage hour in the week.
3. Write the proposal the same day as the call. Everything that makes it convert is in the client's own phrasing, and that decays within hours.

Fable 5.1 builds the system. Prosp delivers the audits and books the calls 💜

↳ Start a 14 Day Free Trial 💜 ↳ Watch Full Tutorial ▶️