Fractional CMO scaling past 3 retainers without becoming an agency

Your three highest-leverage deliverables produce a 60% better first draft, in 40% less time

The monthly client review, the quarterly plan, and the campaign brief are the three artifacts every client engagement orbits around. Each one currently takes 3–5 hours of your direct time per client per cycle — call it 12 hours/week on prep + drafts across three retainers. The outcome: a system where YOU still write the strategic spine (the unique judgment), but the first draft assembly, source-quote pulling, and structural scaffolding happen via prompts and templates you built once. Quality goes up because the draft you sit down with already includes the prep you used to do mentally. Time goes down because you stop re-inventing structure on every artifact.

Why this matters: This is the single change that unlocks taking client #4 without dropping client #2. 12 hours/week recovered ≈ one full client's worth of capacity. Without it the cap stays at 3 retainers and $310K.

How they'll know it's real: Within 90 days, the monthly review for at least 2 of 3 clients is delivered in ≤2 hours of your time (down from 4-5), and the client reports it's better than the prior cycle's — unprompted.

Because you said: Catalyst · Buyer Outcome · Buyer

Your prep before client conversations is deeper than it was — without taking more of your time

Right now your prep is ~30 minutes of skimming the client's Notion, last meeting's notes, and the last week's metrics. After: a 5-minute morning routine where Claude (or equivalent) reads the running client doc, the last meeting transcript via Granola, and any new metrics, and returns a one-page brief: what's changed since last week, what the client is likely to surface, and the two questions you should be ready to answer. You still walk into the meeting as the strategist — but you walk in primed by structured prep, not by improvisation.

Why this matters: Clients hire you for pattern recognition. Better prep = sharper patterns visible = the engagement deepens. Without this the experience drifts toward 'available consultant' instead of 'embedded leader.'

How they'll know it's real: By month 2, at least 1 client mentions you 'spotted something' they hadn't surfaced themselves — and it came directly from your AI-assisted prep, not a coincidence.

Because you said: Buyer Outcome · Buyer · Stakeholders

Your VA runs back-office on a written checklist — you stop being the bottleneck for scheduling, invoicing, and basic research

Today your VA is reactive: she does what you ask, when you ask. That's ~3 hours/week of YOUR time generating requests and reviewing outputs. The outcome: a 1-page weekly checklist she owns end-to-end — Calendly maintenance, monthly invoicing, prospect research enrichment, weekly client-doc tidying. You see results, not requests. The VA's $300/month becomes leveraged through repeatable structure, not your judgment.

Why this matters: F4 says you can't assume a team you don't have. A VA with a checklist IS the team — but only if the checklist exists and works without you. Without this, every back-office task interrupts your client thinking.

How they'll know it's real: By week 6, the VA completes 3 consecutive weeks of the checklist with zero questions to you, and the back-office artifacts (invoices sent, calendar clean, prospects enriched) are visible in your Friday review.

Because you said: Buyer · Constraints · Stakeholders

Client-confidential work stays out of consumer LLMs — and you can prove it to a client who asks

Every client conversation, every client doc, every internal metric is privileged. The outcome: a written and lived rule — client data goes only through tools that don't train on inputs (Claude Pro with project mode, Granola with paid plan, Notion paid). Any tool that DOES train (free ChatGPT, free Gemini, free Notion AI on the consumer plan) is used only with synthetic or sanitized inputs. You have a one-paragraph 'how I use AI in client work' you can paste into a client conversation if asked.

Why this matters: F5 names this as a hard limit. One leak — even theoretical — and a Series A board can ask you to walk. The methodology has to be defensible before it's efficient.

How they'll know it's real: By week 4, you have the written rule pinned in your own Notion + a client-facing one-paragraph version ready. By month 3, you've actually used the client-facing version (a client asked).

Because you said: Constraints · Stakeholders

Your IP compounds — the next client engagement is meaningfully faster and sharper than the last

Today, each client starts roughly from zero: you have your patterns in your head, but the artifacts (prompts, templates, brief structures) live scattered in old docs. The outcome: a 'practice library' of versioned prompts and templates indexed by deliverable type and client archetype — the monthly review prompt at v3 for Series A SaaS, the campaign brief template at v2 for B2C extensions. Each engagement contributes one or two refinements; nothing is rebuilt from scratch.

Why this matters: This is the bedrock of your north star — 'asset-not-time' compounding. Without it, your practice is just your hours, and 32 hours/week caps revenue at the rate. With it, $450K at 32 hours is the same hourly load delivering more value because the assets do the assembly.

How they'll know it's real: By month 6, when you sign client #4, the first month of work uses 50%+ existing assets pulled from the practice library — and you spend the recovered time on her unique strategic problem, not on rebuilding deliverable structure.

Because you said: Identity · Buyer Outcome · North Star

Physical truth — one person, ~32 client-facing hours/week, three clients, one VA

Mon = prep, Tue/Wed/Thu = client work (mornings) + writing/dev/research (afternoons), Fri = your own business (positioning, sales, the LinkedIn essay). VA available ~6 hrs/week, contractor, async. No physical office. Single laptop. No second seat for anything — if a tool requires teammate licenses, it doesn't fit. Single parent: childcare ends at 5:30pm sharp; nothing can require you after that, period. School-year cadence shapes everything (the 'spend a weekend on it' anti-pattern is real).

Why this matters: Every workflow design has to survive these. A 'great' design that quietly assumes a Saturday morning to maintain is a fragile design — it dies in October when school activities ramp up.

Because you said: Buyer · Constraints

Configured truth — current tool stack is patchwork, not ecosystem

Today: ChatGPT (free), Granola (paid solo), Notion (paid solo), Calendly (paid solo), Google Workspace (paid), one shared Google Drive per client. ~$130/mo total recurring. Tools work but nothing flows between them — every artifact is copy/pasted manually. No prompt library; you re-write the same kind of prompt from scratch each time. No client-specific 'project' boundaries inside any AI tool — everything goes into ChatGPT's open chat with no project memory. No backup for the practice library because the practice library doesn't yet exist as a concept.

Why this matters: $310K of revenue is currently held together by your memory. A laptop crash today loses six months of practice tacit knowledge. The configured truth has to catch up to the value the practice produces.

Because you said: Constraints

Operational truth — three client engagements running, one VA, no documented internal process

Each client engagement runs on memory: standing weekly call with each, monthly written review, quarterly plan, ad-hoc Slack between. You hold the cadence in your head and it works — but only because you only have three. With four, two would slip the same week sometimes. VA tasks land in Slack DMs to her and get done; nothing is documented enough that a second VA could pick up tomorrow. The 'no agency' identity makes this an explicit choice — but it means YOU are the process, and the process can't compound.

Why this matters: The asset-not-time outcome (F7) demands that some of what's currently in your head become artifacts. Not all — your unique judgment IS the value — but the recurring container of the engagement can be a template you re-fill, not a thing you re-invent.

Because you said: Identity · Buyer Outcome · Stakeholders

Financial truth — $310K revenue at ~3 retainers, no recurring tool spend headroom needed, $400-600/mo new budget available

Three retainers averaging $10K/mo. Annual: $310K (factoring December slowdown + occasional ad-hoc projects). Solo P&L: ~$30K/year operating costs (VA $300/mo, current tools $130/mo, accountant, software dev fees, insurance). Net before tax ~$280K. Comfortable but single-parent + household-sole-income means it has to stay >$250K through any transition. You can spend $400-600/mo recurring on new tools if it's clearly load-bearing; $1500 one-time if 90-day payback. NOT $5K/mo on Salesforce or HubSpot Enterprise.

Why this matters: Any recommendation that pushes recurring above $1,000/mo OR that requires a $2K+ one-time setup violates the budget. The methodology should bias toward tools with strong $20-$80/seat solo tiers, not enterprise plans designed for teams of 20.

Because you said: Buyer · Constraints

Strategic truth — your moat is judgment shape, NOT delivery throughput

Why clients pay you $10K/mo: you've BEEN a VP Marketing twice, you can spot what's stuck in 30 minutes, and your prescription is opinionated not encyclopedic. A junior person with the same playbook can't replicate this — the recognition is path-dependent. The strategic risk: AI tools used carelessly can flatten you toward 'consultant who runs the AI workflow' instead of 'leader who uses AI to deepen judgment.' The configured truth above is the visible version of this — the same care that protected your moat for 18 months has to extend into how you bring AI in.

Why this matters: Every tool/prompt decision later in this binder should be filtered through: does this DEEPEN judgment (good) or REPLACE judgment with workflow (bad)? That's the single defensible filter that protects the asset-not-time outcome.

Because you said: Identity · North Star

Decision: Build one ‘thinking partner’ workspace per client — not one shared chat

ChatGPT free or a single open Claude chat is the wrong container — context bleeds, security is undisclosed, and you can't trust where prior client info goes. Decision: Claude Pro with Projects (or ChatGPT Team with workspaces) where each retainer client gets its OWN project with its OWN instructions and document set. Adding a new client = duplicating the project template, not starting over.

Why this matters: Solves the confidentiality hard limit and the asset-compounding outcome in one move. Each project becomes its own compounding container.

Because you said: Constraints · North Star

Decision: The three highest-leverage deliverables (monthly review, quarterly plan, campaign brief) become versioned templates with a starter prompt — NOT freeform every time

These three artifacts represent ~60% of recurring deliverable time across all clients. They go into your practice library as Notion templates with a partnered Claude project prompt. The TEMPLATE is the structure (what sections, what order, what questions). The PROMPT is what you paste alongside this engagement's notes to fill in the first draft. Each engagement contributes one refinement — v1 → v2 → v3.

Why this matters: This is the actual mechanism of asset-not-time compounding. The first time costs 4 hours; the fourth time costs 1.5 hours and is better.

Because you said: Buyer Outcome · North Star

Decision: Granola is the meeting layer — not Otter, not Fireflies

Granola sits silently in your menu bar, captures audio + your live typed notes, and produces a structured summary you can ask further questions of. It fits the no-bot-joining-meetings preference (clients hate Otter showing up) and the privileged-conversation constraint (Granola has a no-train commitment on its paid tier). The decision is to STOP using any other note tool — one container, one habit.

Why this matters: Meeting notes are the single largest time leak after deliverable drafting. Replacing your hand-typed notes with Granola + your prompt for post-meeting structuring saves ~3 hours/week and improves recall.

Because you said: Buyer · Constraints · Stakeholders

Decision: Build the practice library in Notion (not Obsidian, not Google Docs)

You already pay for Notion, your clients use Notion (sharing is friction-free), and Notion AI can query across pages within a workspace. The library lives at /practice-library/ with sub-pages for /prompts, /templates, /voice-corpus, /case-stories. Each item is versioned in the title (e.g., monthly-review-prompt-v3). The decision is explicit because the temptation to ‘try Obsidian for personal IP’ would split the library across two homes — and a split library compounds zero things.

Why this matters: One library, one home, one habit. The compounding outcome dies the moment you have two homes for IP.

Because you said: Constraints · North Star

Decision: The VA's work is documented as a 1-page weekly checklist, not as ad-hoc Slack

Today you generate VA work via Slack DMs — high context for you, zero leverage for her. Decision: one Notion page titled ‘VA weekly checklist’ with a fixed daily and weekly rhythm. New tasks get ADDED to the checklist, not DMed. Anything that recurs more than 3 times moves into the checklist permanently. Anything escalation-level routes to a 24-hour async channel, not real-time.

Why this matters: Solves the ‘nothing can assume a team you don't have’ constraint by making the VA effectively that team — for the back-office. Without this, your VA hours stay capped at what you can request in real time.

Because you said: Buyer · Constraints · Stakeholders

Decision: Perplexity (or Claude.ai web search) for client-facing research, NEVER consumer ChatGPT

Client research — competitor positioning, category trend, win/loss-style synthesis — gets routed through a tool with citations you can verify and a paid tier that doesn't train on inputs. Perplexity Pro is the working choice. Free ChatGPT for client research is OFF the table because (a) it trains on inputs and (b) hallucinations without citation chains can't be trusted in a client deliverable.

Why this matters: Protects the confidentiality limit AND the moat (clients catch you cite-checking AI output once and your judgment-credibility erodes).

Because you said: Constraints · Stakeholders

Decision: NO custom API code, NO self-hosted infrastructure, NO bespoke automation

Make.com / Zapier ONLY where they're stupid-simple and the VA can read what's happening. The ‘consultant who became a tools person’ failure mode is real and would actively destroy the moat. If a workflow needs a developer, it doesn't get built — until the practice scales past one person, at which point that decision gets re-opened. Today, every workflow has to be readable by Maya AND by the VA.

Why this matters: This is the no-code hard limit captured as an active filter. It eliminates ~70% of the AI-tools-for-consultants advice on the internet (most of which assumes you're a builder) and is what makes the rest applicable.

Because you said: Buyer · Constraints

Decision: One Friday afternoon per quarter is reserved for ‘practice-library hygiene’ — non-negotiable

Without a scheduled refresh moment, the library decays. Decision: Q1/Q2/Q3/Q4 last Friday afternoons are reserved for: review the prompts/templates touched that quarter, version-bump anything refined, retire anything obsolete, write one ‘what shifted in my judgment this quarter’ note for the voice corpus. This is the mechanism that turns the asset into an asset, not a graveyard.

Why this matters: Closes the operational gap between the asset existing and the asset compounding. Without it, six months from now you'll have 40 unversioned prompts in scattered files and no memory of which one is current.

Because you said: Buyer Outcome · North Star

Per-client thinking partner with project-scoped memory

One AI workspace per retainer client, with project instructions, the client's running doc, and the latest meeting transcripts loaded. You open the project, ask one question, and the answer is already shaped by what THIS client said three weeks ago — not a generic answer. Recommended via: Claude Pro with Projects (privileged-by-default, supports docs in project memory).

Why this matters: Without per-client scoping, every conversation starts from zero. With it, the AI becomes a second junior strategist who's been in every meeting.

Because you said: Buyer Outcome · Constraints

Meeting capture → structured summary in your voice

Granola sits in the menu bar, records audio, captures your live notes, then on demand produces: a structured summary, a decisions list, a 1-line ‘what shifted’ note. You ask follow-up questions of the transcript without re-listening. The summary lands in the client's Notion the same day, formatted to your template.

Why this matters: Meeting note-taking is the largest single time leak in your week (~3 hrs across all clients). This integration recovers most of it without changing your behavior in the room.

Because you said: Buyer Outcome · Buyer · Stakeholders

Versioned prompts for the three recurring deliverables (monthly review, quarterly plan, campaign brief)

Three prompts live in the practice library, each pinned at the current version. You paste this engagement's context (last month's review + this month's metrics + the meeting transcripts) and the prompt produces a structured first draft — sections in the right order, the right questions surfaced, your voice. You write the strategic spine; the assembly is done. Each quarter you version-bump anything you refined.

Why this matters: Closes the 60% time-on-deliverables outcome directly. Builds the practice library piece by piece as a side-effect.

Because you said: Buyer Outcome · North Star

Pre-meeting prep brief — generated from running context, every Tuesday morning

Each client has a standing weekly call. A prep prompt run against the client's project produces (in 5 minutes): what's changed since last call, the two things this client is likely to surface, the one question you should be ready to answer. You skim it, mark up what's wrong, walk into the call primed.

Why this matters: This is the integration that closes the ‘deeper prep without more time’ outcome. The signal of success is the client noticing depth they didn't bring up themselves.

Because you said: Buyer Outcome · Stakeholders

Client-research with citations — never bare-LLM hallucinations

All client-facing research (competitor positioning, category trend, win/loss synthesis) goes through Perplexity Pro or Claude with web-search enabled, with citations visible. You spot-check the top citation before the synthesis lands in a deliverable. Bare-ChatGPT-without-citations is out of scope for client work.

Why this matters: Protects judgment-credibility. One time a client catches a hallucinated stat in your deliverable and the moat erodes.

Because you said: Constraints · Stakeholders

VA-readable back-office checklist with weekly cadence

A Notion page the VA owns. Daily section (calendar maintenance, inbox triage rules), weekly section (invoicing, prospect-research enrichment, client-doc tidying), monthly section (financial reconciliation, prospect-pipeline review). You see results in a Friday review, not requests during the week.

Why this matters: Solves the bottleneck where YOU generate VA work. Without it, the VA's hours scale with your real-time availability, not with the actual work to be done.

Because you said: Buyer · Stakeholders

Voice corpus — your writing samples and judgment samples, indexed

A growing collection of your best LinkedIn essays, client memos, and a quarterly ‘what shifted in my judgment’ note. Loaded into Claude project instructions so every prompt-generated draft sounds like Maya, not like a default AI voice. The corpus is the secret sauce that makes the prompts produce output you'd publish.

Why this matters: Voice is the most copy-pastable piece of your moat. Without it, every AI-assisted draft requires heavy re-writing because it sounds wrong.

Because you said: Identity · Buyer · North Star

Sanitized-input rule for any tool that trains on data

Hard policy: if a tool's terms allow training on inputs, you only use it with sanitized inputs — no client name, no metrics, no proprietary positioning. Anonymized scenarios, generic context, hypothetical numbers. The rule is written into a 1-paragraph internal policy AND into a client-facing one-paragraph version you can paste into a conversation.

Why this matters: Defensible. A client can ask you tomorrow how you use AI; you have an answer that doesn't require a lawyer to read.

Because you said: Constraints · Stakeholders

Calendar-guard integration — protected client/writing/admin blocks

Reclaim.ai or similar light tool that automatically defends Tue/Wed/Thu mornings as client-only, Mondays as prep, Fridays as your-own-business. Auto-declines meeting requests in the wrong blocks. Surfaces to prospects via Calendly that you have specific windows, not ‘pick anything.’

Why this matters: The single-parent + 5:30pm-hard-stop reality requires the calendar to defend itself, because you won't always have the energy to defend it manually.

Because you said: Buyer · Constraints

Quarterly practice-library refresh — the asset-compounding ritual

Last Friday of each quarter, 3 hours: version-bump refined prompts, retire stale ones, write the quarterly ‘what shifted in my judgment’ note, refresh the voice corpus, archive one client artifact per client that taught you something. This is not a tool — it's a recurring integration of YOUR time with the library, scheduled in advance.

Why this matters: An un-refreshed library decays into a graveyard within two quarters. The compounding outcome dies the moment the refresh stops happening.

Because you said: Buyer Outcome · North Star

Daily — VA inbox triage + calendar maintenance (15 min, async)

VA runs once each morning. Reviews calendar invites against your protected blocks (Reclaim handles auto-decline; VA handles judgment calls). Triages your inbox using a written rule sheet: client = pass through, prospect = log to pipeline + draft a one-line response for your approval, vendor/admin = handle directly, everything else = archive. Output: a 3-line end-of-day Slack to you summarizing exceptions only.

Why this matters: Closes the bottleneck where you respond in real time to inbox/calendar drift. Saves ~30 min/day of yours and gives back the start of the morning to client prep.

Because you said: Buyer · Stakeholders

Tue/Wed/Thu morning — 30-minute client prep block (you, per client)

9:00 — open the client's Claude project. Paste this morning's prep prompt. Read the generated brief. Mark up what's wrong. Walk into the 10am call primed. Total: 30 minutes per client per week instead of 30 minutes of skimming with no structure.

Why this matters: This is the operational mechanism of ‘deeper prep without more time.’ Same 30 minutes; the difference is the AI-assisted brief vs. unstructured skimming.

Because you said: Buyer Outcome · Buyer

After every client call — 10-minute post-meeting structuring (you)

While Granola is still warm, ask it to produce: structured summary, decisions list, your verbatim commitments, the one ‘what shifted’ note. Paste the result into the client's Notion. Total 10 minutes. The transcript itself becomes archival — you work from the structured artifact, not the raw recording.

Why this matters: Without this routine, meeting notes pile up untouched and a week later you're reconstructing from memory. With it, every meeting produces a durable artifact same-day.

Because you said: Buyer Outcome · Stakeholders

Monthly per-client — monthly review production (you + Claude project, ~2 hrs)

Last business day of each calendar month: open the client's Claude project, paste this month's metrics + the structured summaries from each weekly call. Run the monthly-review prompt (vN). Edit the strategic spine. Polish. Deliver. Replaces the current 4-5 hour version. The 'edit the strategic spine' is where YOUR judgment goes; everything else is assembly.

Why this matters: Single largest deliverable, three times a month, every month. Going from 4-5 hrs → 2 hrs per client = 6-9 hours/month recovered.

Because you said: Buyer Outcome · North Star

Weekly — VA back-office sweep (Thursday afternoon, ~2 hrs)

VA runs the weekly checklist Thursday afternoon: send any invoices due, enrich the prospect pipeline (one-pager research per new lead in the queue), tidy each client's running Notion (link in this week's artifacts, archive last week's), refresh the Friday review packet for you. Output: Friday morning, you open the packet and it's current.

Why this matters: Compresses 5+ hours of scattered back-office into one block, run async, on a checklist. No more Thursday-evening ‘oh no, I forgot to invoice’ moments.

Because you said: Buyer · Constraints

Friday morning — your-business block, including a 30-min library-maintenance habit (you)

9:00-12:00. The 30 minutes is for: did I refine any prompt or template this week? Bump its version, add a note. Did I write something publishable? Drop it in the voice corpus. Did a client teach me something? Write the one-line ‘what shifted.’ The rest of Friday morning is positioning + sales + the weekly LinkedIn essay.

Why this matters: The library decays without weekly micro-attention. 30 minutes of attention is enough to keep it alive; missing two weeks in a row puts it on the path toward graveyard.

Because you said: Buyer Outcome · North Star

Quarterly — Practice-library refresh (last Friday afternoon of the quarter, 3 hrs)

The deeper version of the weekly habit, plus: review which prompts produced the best client artifacts this quarter, retire any prompts unused for two quarters, refresh the voice corpus with the quarter's strongest writing, write the ‘what shifted in my judgment this quarter’ note. Output: a one-page quarterly memo to yourself that you'll re-read at the next refresh.

Why this matters: This is the ritual that converts a library-as-pile into a library-as-asset. Without it, six quarters from now you have hundreds of artifacts and no way to find the right one.

Because you said: Buyer Outcome · North Star

Escalation rule — anything client-confidential never goes to a training-enabled tool, no exceptions

If you catch yourself about to paste a client name, a real metric, or a proprietary positioning into ChatGPT free or any free-tier AI, STOP. Either route it through Claude Pro project (privileged), Perplexity Pro (no-train on Pro), or sanitize the input. If sanitizing kills the value, you write the section unaided — that's a fair trade for the trust.

Why this matters: The trust IS the moat. The rule has to be in your operational vocabulary, not just in a policy doc — because the moment of decision is at the paste, not in the file.

Because you said: Constraints · Stakeholders

What you're building, in one sentence

A solo fractional CMO practice where AI is your second junior strategist (per-client, privileged, in your voice) and a versioned library of assets (prompts, templates, voice corpus) that compounds across engagements — so the practice at 24 months delivers $450K with 32 client-facing hours/week, NOT by becoming an agency.

Because you said: Identity · North Star

The single most important shift

From ‘consultant who improvises every deliverable’ to ‘consultant whose deliverables are produced by versioned templates filled with this client's specifics, where my judgment is the strategic spine and the assembly is automated.’ Every other decision in the binder serves this one shift.

Because you said: Buyer Outcome · North Star

The shape in months and money

Month 0: $310K/year, 3 retainers, 12 hrs/week leaking on prep + drafts. Month 3: same revenue, 6-7 hrs/week recovered, three core deliverables on v2 templates, VA checklist working. Month 6: client #4 onboarded WITHOUT dropping clients 1-3, revenue trending toward $400K, practice library indexed and searchable. Month 12: $400-430K, 32 client-facing hours/week, library on v4-v6 of core templates. Month 24: $450K, same hours, library is a multi-quarter asset, no agency drift.

Because you said: Catalyst · Buyer Outcome · North Star

What proves this is working

Three signals, in order of when they'd appear: (1) Within 60 days, at least one client says ‘that monthly review was your sharpest one yet’ — unprompted — and you got there in 2 hours, not 5. (2) Within 90 days, three consecutive weeks the VA runs the checklist with zero questions to you. (3) Within 6 months, client #4 onboards and you can show the practice library as the reason your first month with her was 50% pulled-from-shelf, not built-from-scratch.

Because you said: Buyer Outcome · Stakeholders · North Star

The failure mode to actively guard against

‘Becoming a tools person.’ The seductive failure mode is: you start using AI everywhere, the deliverables get faster, but the strategic spine atrophies because the AI is doing more of the thinking and you're doing more of the workflow. The signal: a client engagement where you can't articulate, in plain English, what unique judgment you brought that the AI couldn't have produced alone. If you ever feel that, retire the prompt for that deliverable for a quarter, write it by hand, and rebuild the prompt from what you wrote.

Because you said: Identity · Constraints · North Star

What stays out of scope on purpose

(a) No agency hires — no junior consultants, no account managers, no delivery team. (b) No custom code or self-hosted infrastructure. (c) No enterprise-tier tool spend (>$500/mo per tool). (d) No multi-tool publishing automation — your LinkedIn essay is hand-written, full stop. (e) No client-data flow into open consumer LLMs, ever. Each of these would either violate the asset-not-time bedrock or the confidentiality limit.

Because you said: Buyer · Constraints · North Star

Week 1 — Stand up Claude Pro with one Project per retainer client

WHY: Outcome #1 (better drafts in less time) and outcome #4 (privileged client work) both require a per-client context home that isn't an open consumer chat. WHERE: claude.ai (Pro, $20/mo). WHAT: 1) Sign up for Pro, 2) Create one Project per client (Client A, Client B, Client C), 3) Paste a 1-paragraph 'project instructions' for each — their name, what they sell, their stage, what's stuck, the agreed quarterly focus, 4) Upload the last two months of meeting summaries + the running client doc to each project. GOTCHAS: The temptation is to load EVERYTHING — old presentations from 2024, every Slack export. Don't. Load only what's still operative; old context confuses. VERIFICATION: Open Client A's project, ask 'what's the single thing most worth doing this quarter for [Client A]?' and the answer references this quarter's actual focus, not generic SaaS advice.

Why this matters: This is the foundation. Every later AI move plugs into these projects. If the per-client scoping isn't there, week 2's prompts can't compound.

How they'll know it's real: Three projects exist, each with project instructions + ≥2 months of context loaded, and a probe question returns a client-specific answer.

Because you said: Identity · Buyer Outcome · Constraints

Week 1 — Move Granola from menu-bar afterthought to first-class capture

WHY: Meetings are the largest single time leak after deliverables. Granola already sits there but you don't use the post-meeting summary feature consistently. WHERE: Granola desktop app + Notion. WHAT: 1) Upgrade Granola to the paid solo tier ($14/mo) so the no-train guarantee applies, 2) Add a post-meeting habit: before closing the laptop after any client call, ask Granola to produce structured summary + decisions + your commitments, 3) Paste the structured summary into the client's Notion immediately, 4) Don't keep the raw transcript past the week unless it had something genuinely re-listenable. GOTCHAS: The 10-minute post-meeting step gets skipped under back-to-back pressure. Recovery: schedule it as a 10-min calendar event tacked onto every client meeting, not as an aspiration. VERIFICATION: Every client call in week 1 has a structured Notion entry within 30 minutes of the meeting ending.

Why this matters: Recovers ~2-3 hrs/week of meeting-note labor and creates the durable artifact future deliverable prompts will pull from.

How they'll know it's real: All client calls week-1 produce a same-day structured Notion entry.

Because you said: Buyer Outcome · Buyer

Week 1 — Write the first version of the monthly-review prompt

WHY: This is the single largest recurring deliverable (3 clients × ~4 hrs = 12 hrs/month). Even a v1 prompt cuts it noticeably. WHERE: In each client's Claude Project, as a saved prompt or a Notion library entry. WHAT: 1) Open the most recent monthly review you wrote (your best one for any client), 2) Write a prompt that, given last month's metrics + meeting summaries + client doc state, produces a draft in the same structure, 3) Test it on next month's review for Client A first — read what it produces, mark up what's wrong, refine the prompt, 4) Save as monthly-review-prompt-v1 in the practice library Notion. GOTCHAS: Don't try to make v1 perfect. v1 is the worst it'll ever be; you'll bump to v2 after using it once. VERIFICATION: Client A's next monthly review takes ≤3 hours of your time (down from 4-5) and the client reaction is neutral-or-better.

Why this matters: Builds the first concrete piece of the practice library. Once one of these exists, the others are easier — the format is established.

How they'll know it's real: monthly-review-prompt-v1 exists in the library AND has been used on a real client deliverable.

Because you said: Buyer Outcome · North Star

Week 2 — Build the VA weekly checklist in Notion

WHY: Outcome #3 (VA runs back-office from a checklist). WHERE: One Notion page titled 'VA weekly checklist', shared with the VA. WHAT: 1) List everything you've Slacked the VA in the last month, group by recurrence (daily/weekly/monthly), 2) Write each as a one-line task with the explicit rule (e.g., 'Send Calendly invoice on the 1st using the template at /templates/invoice-v1; if amount > $5K, ping me for approval'), 3) Add an 'exceptions to escalate' section with 3-5 rules (e.g., 'any inbound from someone with a title containing CEO or Founder = escalate same day'), 4) Walk through one pass with the VA on a Tuesday afternoon call. GOTCHAS: First two weeks the VA will ask questions in real time out of caution — route them to a 24-hr async channel. VERIFICATION: VA completes one full week using only the checklist (no Slack DMs from you generating tasks).

Why this matters: Converts the VA from a reactive contractor into a routinely-running back-office function. Without this, the VA's leverage caps at your real-time availability.

How they'll know it's real: One clean week of VA running checklist with zero same-day questions to you.

Because you said: Buyer · Constraints · Stakeholders

Week 2 — Add Perplexity Pro for client-facing research

WHY: Client research that lands in deliverables needs citations you can verify. Free ChatGPT for client research is off the table (training + hallucination). WHERE: Perplexity Pro ($20/mo). WHAT: 1) Sign up for Pro (the no-train commitment applies), 2) On your next client-research need (competitor positioning, category trend, win/loss synthesis), use Perplexity instead of any consumer LLM, 3) Spot-check the top citation before any synthesis lands in a deliverable, 4) If Perplexity's answer is thinner than you expected, that's the signal to pull primary sources yourself — don't paper it over with a longer prompt. GOTCHAS: The Pro tier's no-train applies to Pro inputs only — don't run client research on the free tier 'to save the seat.' VERIFICATION: One client deliverable in week 2 includes research backed by Perplexity citations you've verified.

Why this matters: Protects judgment-credibility (citations) AND the confidentiality rule (paid tier) in one move.

How they'll know it's real: Perplexity Pro in use AND visible in one client deliverable that week.

Because you said: Constraints · Stakeholders

Week 2 — Set up Reclaim.ai to defend the client/admin/personal blocks

WHY: Single-parent + 5:30pm hard stop + the Tue/Wed/Thu morning client structure can't be defended by manual calendar management. WHERE: Reclaim.ai (free tier is enough; paid is ~$8/mo). WHAT: 1) Connect Google Calendar, 2) Define recurring 'protected' blocks: Tue/Wed/Thu 9am-1pm = client work, Mon 9am-5pm = prep, Fri 9am-5pm = own business, 3) Set 'no meetings' policy on protected blocks; auto-decline external requests, 4) Connect to Calendly so external bookings can only land in your defined 'meeting' windows. GOTCHAS: Auto-decline can offend a real client if it bounces something urgent. Build a 'priority' rule: requests from your 3 current client domains bypass the auto-decline and ping you instead. VERIFICATION: One week passes where no meeting requests land in a protected block.

Why this matters: The calendar guarding has to be passive; if you have to defend it manually every week, you won't (especially during a school-year ramp).

How they'll know it's real: One clean week of protected blocks unviolated by external meeting requests.

Because you said: Buyer · Constraints

Week 3 — Write the campaign-brief and quarterly-plan prompts (v1 each)

WHY: Same pattern as the monthly review — these two deliverables are the next-largest recurring time leaks. WHERE: Practice library Notion. WHAT: 1) For each, find the strongest example you've written in the last 12 months, 2) Write a prompt that, given the client context + the unique campaign/quarter context, produces a draft in your structure, 3) Test each one on a real upcoming deliverable, 4) Save as campaign-brief-prompt-v1 and quarterly-plan-prompt-v1. GOTCHAS: The quarterly plan especially has more strategic spine than structure — a prompt that flattens that into a checklist is worse than no prompt. If v1 produces something generic, the prompt is wrong; refine the project instructions, not the prompt. VERIFICATION: Both prompts exist, both have been used on one real artifact each.

Why this matters: Completes the three-deliverable foundation. After this week, all three highest-leverage recurring artifacts have a v1 in the library.

How they'll know it's real: Three deliverable prompts in the library, each used on at least one real artifact.

Because you said: Buyer Outcome · North Star

Week 3 — Seed the voice corpus with 8-12 samples

WHY: Without voice samples in the project instructions, the AI drafts sound like default-AI, not like Maya. WHERE: A single Notion page in the practice library, /voice-corpus/. WHAT: 1) Pull your 4-5 strongest LinkedIn essays from the last 6 months, 2) Pull 3-4 client memos or quarterly plans that read as YOU at your sharpest, 3) Add a one-paragraph 'how I write' note: what you do (specifics over generics, opinion before disclaimer, lead with the contradicting thing), what you don't (em-dashes everywhere, bullet-soup, AI-tells like 'In today's landscape'), 4) Add 'load the voice corpus into the project instructions' to each Claude project's setup. GOTCHAS: Voice samples don't survive vague — if you can't say in one sentence what makes a sample 'you,' it's not going to make the AI sound like you. VERIFICATION: A draft from any client project, with voice corpus loaded, reads as YOU at first scan, not as default AI.

Why this matters: Voice is the most copy-pastable piece of your moat. Without samples, every draft needs heavy re-writing — defeating the time-recovery outcome.

How they'll know it's real: Voice corpus has 8-12 entries, loaded into all three client projects.

Because you said: Identity · Buyer · North Star

Week 4 — Write the client-facing AI-use policy (1 paragraph) and pin it

WHY: Outcome #4 (provable confidentiality). Clients will ask; you need a non-defensive answer ready. WHERE: Practice library + as a saved Notion template you can paste into a client conversation. WHAT: 1) Write one paragraph that names: which tools you use (Claude Pro projects, Granola paid, Perplexity Pro), the no-train commitments those tools make, the categories of work where you sanitize inputs (any consumer-tier tool), and what you DON'T use AI for (the strategic spine of any deliverable), 2) Have a colleague who is a client-side marketing leader read it and tell you what would worry them, 3) Refine, 4) Pin it. GOTCHAS: A bullet list of tools is the wrong format — clients want to know you've thought about the principle, not just shopped for vendors. VERIFICATION: One non-client trusted reader can summarize, in their own words, what you do and don't do — without reading it twice.

Why this matters: The first time a client asks (and one will), having this ready turns a moment of risk into a moment of trust-deepening. Without it, your answer drifts toward defensive.

How they'll know it's real: Policy paragraph written, reviewed by one trusted external reader, pinned in the practice library.

Because you said: Constraints · Stakeholders

Week 4 — Schedule the Quarter-1 practice-library refresh + the day-90 checkpoint

WHY: The asset-compounding outcome dies without scheduled refresh moments. WHERE: Your own calendar. WHAT: 1) Block the last Friday afternoon of the next quarter as 'Practice library refresh — non-negotiable' (3 hrs), 2) Block a day-90 checkpoint at 2 hrs to honestly answer: are the three deliverable prompts saving the time projected? Is the VA running the checklist solo? Did I keep client data out of consumer LLMs all quarter? Did I add or refine at least 4 library items? 3) Build a one-page checkpoint template now while the questions are fresh. GOTCHAS: You will be tempted to skip both because 'this quarter is busy.' Recovery: the calendar blocks are the same Friday afternoon slot, just doubled — you're not adding time, you're using existing time. VERIFICATION: Calendar invite exists for both, template exists for the day-90 review.

Why this matters: Without a scheduled honest re-read, you'll discover in month 5 that the library is decayed and the VA is back to ad-hoc Slack. The mechanism only works if the mechanism is scheduled.

How they'll know it's real: Calendar blocks exist, template exists.

Because you said: Buyer Outcome · North Star