The 100-Call Machine

Rapid bootstrapped growth · standing target until Tom calls capacity

The Target, Decomposed

100 booked a month = 25 a week = about 4.7 per working day (Israel Sun-Thu, about 22 working days). At the engineered 75% held-rate: 75 held calls a month. This table is the only arithmetic that matters; every other section on this page exists to make one of these rows true.

The Named Call Offer

The name is decided; it goes everywhere from week 1. The binary framing wins because the visitor pre-imagines both outcomes and both are valuable. "Not worth building" is already the best sentence on the site; naming the call after it promotes it from FAQ to offer.

The promise line, verbatim, for the /book hero, DMs, and the one-pager footer. Micro-note under it: Free · 10 minutes · with the builder · no obligation.

The risk-reversal stack (publish on /book + one-pager footer)

Never promise "no pitch" (untrue). Instead: "the only thing offered on this call is a next step whose price is already public."

The artifact-first cold offer: the One-Page Technical Read («קריאה טכנית בעמוד אחד»)

The single highest-leverage asset in the plan. It is the only cold offer that works inside Amendment 40 at scale, and the substitute for the testimonials this business is not yet allowed to have.

• New /read page. Name, phone or email, one textarea (3-6 sentences on the idea or the process), optional link. Promise: a personalized one-page technical read within two business days. Verdict / Approach / Cost band / First step / One risk we'd check. Free. No call required.
• Production line. n8n intake → agent research and draft against the fixed 5-block template → Tom reviews about 10 minutes → sent as a branded mockup URL in the thread the requester opened. Capacity: 6-8 a day at about 1 hour of Tom review total.
• Conversion to call. Every read ends with 2-3 open questions specific to their case, plus the handover line below.
• As cold outreach it IS the offer. Offered 1:1, delivered on yes. Never blasted.

What changes on /book (ships week 1)

• Named offer + the three-answer promise as the hero. About 12 string locations in LanguageContext.jsx + extraStrings.js, he/en.
• Live slot count rendered from the existing availability fetch. Rendered, never paraphrased: it cannot lie. This is the only scarcity in the plan and it is real.
• Proof-of-work band (the social-proof substitute, all true): three demo thumbnails linking to /demos ("10 working systems you can click right now"); the production line "A system built here price-checks 1,121 products across 8 sites every night and runs 10 alerting jobs. The client stays unnamed."; a link to the sample viability assessment; and "The call is with the developer who would build it."
• Auto-confirm on shown slots (built in the sprint, section 05). Flagged clearly there: it reverses the deliberate 2026-08-03 request-to-approve design and it is Tom's call.
• 90-second explainer, Tom on camera: what happens in the ten minutes, 20 seconds of clicking a live demo, "sometimes the answer is don't build". Recorded once, he+en subtitles, reused as the DM attachment.
• ₪250 fast-action variant one rung down. No discounts, ever; the fee becomes a deposit.

Every Morning, Agents Have Already Delivered

By 08:00, unattended and overnight, the agent team puts eight things on the table. Tom's hours go only into send, dial, and talk: the three acts the law and the platforms require to be human.

Tom's day (Sun-Thu; Operator pings at each block start)

Five Engines, With Ramp

Sequence engineered around the chagim: front-load what works while operating businesses are absent. Day 1: warm blitz + declared-intent + content + LinkedIn warming. Wk 3: phone (Q4-prep verticals) + partner recruiting. Wk 4: paid. Oct 5: phone expands to all verticals and the September pre-books land.

Speed-to-Lead + Show-Rate Machine

5.1 Kill the approve gate (0.5-1 day)

Extract Create Event + Mail Lead Confirmation from funnel-booking-approval.json into a shared sub-workflow. In funnel-book.json, after Insert Hold + the Notion upsert, branch: IF the slot is inside approved blocks AND it is the first booking from this contact AND the email domain is non-disposable AND there is no prior no-show flag → confirm instantly, with WA Ping Tom becoming an FYI rather than a gate. Everything else falls to today's manual approve page unchanged. Guardrails replacing the human: max 1 active booking per phone+email (holds-table dedupe), max 3 consecutive slots without a gap (Candidates node), and a WhatsApp kill switch that re-enables the gate globally. Site SLA copy goes from "confirmed within 1 business day" to "confirmed instantly." Worth roughly 10-15 booked and held calls a month at target volume.

5.2 The reminder ladder

Operator + reminder_dispatcher.py; templates pre-approved once as cron-class. Unmanaged free calls hold at 50-60%; this ladder plans at 75% (band 70-80), worth about 20 held calls a month. Instrument it, never assume it.

• T0, instant: confirm on WhatsApp + email with a Google Calendar event / .ics attached (the single biggest held-rate lever), plus the micro-commitment question below.
• T-24h: reminder + the pre-call artifact link ("Tom already prepped an initial read of [your thing] and is bringing it to the call").
• T-2h: tap-to-confirm nudge, reschedule-not-cancel framing (template below).
• T-10min: Meet link resent.
• No-show +4 min: one we-missed-you with a one-tap rebook link. A second no-show routes async-only, with the flag stored in the holds table so auto-confirm sees it.

5.3 Pre-call intake

The form grows to exactly 3 fields: one sentence on what you're building, a stage picker, and language. Under 30 seconds; cap it there. On confirmation, Operator opens a disclosed-AI WhatsApp thread. The disclosure is mandatory, and it is the demo. Hard AI boundaries: never masquerades as Tom; no pricing, no scoping, no feasibility promises; instant human handoff on any annoyance signal. Output: a 5-line pre-call card (who / building what / stage / blocker / suggested route + the most relevant of the 10 demos) into the calendar event and WhatsApped to Tom 30 minutes before.

5.4 The post-call machine (10-minute SLA)

Call ends → Tom voice-notes 30-60 seconds in the 5-minute buffer (ingest_voice_memos.py rails exist) → an agent merges note + card + form and drafts the package: a 5-line summary, the recommended route with the reason, one next step, the single most relevant demo, and the tokened ₪250 discovery link (funnel-step2-unlock.json already mints tokens and sends WA links). Packaged as a WhatsApp message plus a branded one-page mockup URL carrying the referral footer below. Tom approves with a one-word "send"; the external-comms gate is preserved. No booking → one nudge at +2 days, one final at +7 with the sample assessment, hard 2-touch cap. This artifact is the ₪250 conversion driver: an estimated 15-19 discoveries a month at 75 held, versus 8-12 without it.

Tom's Dashboard + Dials

This section is not a warning label and it is not a case for slowing down. It is the cockpit: what 100 costs in hours, where the machine physically strains first, and the pre-built relief valves Tom turns when he decides he is at capacity. Until he turns one, the floor stays 100.

The calendar

Sun-Thu, two hard blocks: 11:30-13:00 and 16:30-18:00 Israel time, on a 15-minute grid (10-minute call + 5-minute buffer) = 12 exposed slots a day, 60 a week against about 23 a week needed at target. That is 2.6x headroom, so instant confirm never starves. Implemented in funnel-availability.json: the Candidates node emits only in-block slots and cluster-fills (offering slots adjacent to booked ones first, so calls batch and build windows stay contiguous). During the chagim: one block, 16:30-18:00; founders are the audience still awake. Zero calls Fri/Sat. Blocks bordered by ≥3h protected delivery windows.

Slot-cap decision for Tom now: at target, 5 shown slots a day saturates (~105/mo). Pre-decide the expansion rule (6-7 a day in push weeks) so the displayed scarcity always stays true.

The 10-minute script, by the clock

• Min 0-1: confirm the pre-call card. "I read what you sent. You're building X, blocked on Y, right?"
• Min 1-7: three discovery questions.
• Min 8: the route recommendation.
• Min 9: close + the referral ask.

The calendar event ends at :10 and Meet shows the countdown. Overflow policy: at most 1 call in 5 may eat its buffer; none may eat the next slot. This is load-bearing, not etiquette: every minute of average sprawl adds 75 minutes a month of Tom.

Triage tiers (form-level, before slot exposure; never bounce a booked slot)

Load math

Assumptions to instrument in the Notion Leads-Funnel DB from week 1: 75% held, 20-25% free→₪250, ~35% ₪250→viability. Re-run monthly with real rates.

Funnel 13-15h + two builds 30-35h = 45-50h a week: survivable, zero slack. Where it physically strains, in order: (1) call sprawl, since a 20-minute average doubles the free layer and the timebox is the wall; (2) viability delivery above 8 a month, at 3 unfakeable Tom-hours each; (3) a third concurrent build.

Tom's dials

Pre-built relief valves. Each one is Tom's to turn, never the plan's to turn for him, and none of them touches the 65 or the 100. Note the last row: a setter is not the fix below 150-200 booked a month; the binding constraint at 100 is delivery hours, not dial volume.

Rider, Not Engine

Switch on week 4, after outreach has revealed which hooks land; starting sooner burns budget on message-testing that the manual channels do free. Stand the ad account up now so it is ready (the Meta programmatic layer is already built). Concentrate, never spread.

• 70% (~₪1,400/mo): Hebrew lead-form ads offering the artifact, not the call: the sample viability assessment / one-page read angle, to interest-targeted Israeli founders. The lead form carries an explicit WhatsApp-consent checkbox («אשמח לקבל את המסמך ומעקב בוואטסאפ»), which legally opens Operator nurture per lead under Amendment 40's written-consent requirement. At Israeli doc-opt-in CPLs of ₪10-20: 70-140 consented leads a month → 8-12% nurture-to-book.
• 30% (~₪600/mo): retargeting site visitors and the /demos players, the highest-intent audience the business has, with the production-story creative and a direct call CTA. Retargeted traffic is the 3% who are ready.
• Rules: creative recycled from the best-performing organic posts, refreshed every 2 weeks; kill any ad above ₪25/lead. CFA gate, monthly: per ~25 paid-sourced booked calls, ≥3 buy the ₪250 and ≥1 reaches a viability assessment (≈₪2,500 ≥ spend). At or above the gate → push to the ₪2,000 ceiling. Below it 2 straight months → cut to ₪1,000 and fix creative from organic learnings, never by adding budget.
• Buffer role: if wk 8 tracks below 20 booked a week, this is the headroom lever: the gated-asset boost can add 15-45 calls of slack.

Week by Week to 100

Month 1 at 35-50 is ramp physics, not target-shrinking: cold and content channels warm on a lag. The target fixes the daily activity number; the curve is when each engine reaches temperature. Manage on the leading indicators; they predict bookings 1-2 weeks ahead.

What This Plan Burns, Stated Plainly

20-24 Tom-hours a week on top of 30-35 delivery hours: 45-57 hour weeks, sustainable for a sprint quarter, not forever. All delivery slack (the relief valves in section 06 are the safety margin; Tom turns them at their triggers). Some of the considered-boutique tone at volume, mitigated but not eliminated by every send being Tom-edited and every claim true. And the Monesys/GPI-first priority, which this plan explicitly overrides pending Tom's written decision.

The five real failure modes and their tripwires

Each fix adjusts the machine, never the target.

Legal posture, one line: every 1:1 message is individual, personalized, human-sent; bulk surfaces are content and opted-in ads; email only to opt-ins; phone is human-dialed with DNC screening; nothing automated ever touches Tom's LinkedIn or WhatsApp. Amendment 40 exposure across the whole machine: engineered to zero.

Decisions Tom owes before day 1

1. The priorities.md trade, in writing (boxed above).
2. Auto-confirm reversal of the 2026-08-03 gate (boxed in section 05).
3. The slot-expansion rule.
4. Metered approvals: the job-ad scan cron (a few $/mo), the XPlace membership, Meta at ₪1,500-2,000/mo.
5. The ₪250 refund + 48h credit policy.

Why the Old Ceiling Was 25-40 and This One Is 100

The Verdicts

The plan was pressure-tested through five independent lenses. Condensed verdicts below; each signed off with a named warning.

Tomorrow at 08:00