01 · Thesis

The trusted person for your personal tech — not your IT department

Wealthy, well-connected, non-technical people have a specific gap: nobody in their circle builds custom AI and automation for how they actually work day-to-day. IT companies run cables, manage switches, and handle help-desk tickets. Internal IT keeps the business running. Neither builds you a morning brief that writes itself, an email that drafts in your voice, or a private local AI that keeps your data off the cloud.

That's our niche. We are personalized tech and AI consulting — one-on-one intake, custom automations, concierge delivery — for people who have better things to do than learn new apps.

Where we meet IT

• If a client has internal IT that isn't AI-forward, we collaborate — bring them up to speed, deploy the AI layer, hand them a playbook.
• If a client is scoping a bigger tech deployment, we can help plan it — then hand execution to a real IT partner. That's not our business.

02 · Principals

03 · Target customer

The time-poor operator

• Net worth $5M+
• 40–65, runs a business or carries significant responsibilities
• Does not want to learn new apps — wants outcomes
• Will happily pay $500–2,500/mo to have someone just handle it
• Examples: principals of privately-held businesses, family offices, LobbyIQ clients, Protaea network, senior surgeons / lawyers / consultants

The curious executor

• Tech-adjacent but doesn't want to DIY
• Wants AI in their life without giving data to random SaaS
• Values privacy and bespoke solutions
• Willing to invest $5–20k in hardware for "their own AI"

Not our customer

• Anyone who asks "what's the ROI?"
• Startup founders (they'll DIY)
• Anyone shopping on price
• Anyone who actually needs an IT company — help desk, printer fleets, network overhaul — we refer them out
• Businesses without internal IT who expect us to be their IT department

04 · Service catalog

Five tiers, all organized around personal workflows + custom AI + data. The retainer (Tier A + B) is the core product. Hardware (Tier C) is concierge-grade home infrastructure. Projects (Tier D) are one-off builds. Data & intelligence (Tier E) monetizes our existing scraping stack and LobbyIQ dataset. None of this overlaps with traditional IT work — if a client needs help desk, networking, or Office 365 admin, refer them to an IT partner.

Tier A — "Wow in week one"

• Morning brief — 6am email: overnight inbox summary, today's calendar, top 3 priorities. Passive delivery.
• Smart inbox triage — AI sorts into Respond Today / This Week / FYI / Delete.
• Draft-in-voice for key contacts — spouse, top client, partner. Forward a thread → return a draft.
• Calendar conflict watchdog across personal / work / family / board calendars.
• Birthday & anniversary radar — 2 weeks out with gift suggestions.
• Receipt → expense log — photo to categorized spreadsheet.

Tier B — "Stickiness" (month 2+)

• Follow-up watchdog — "you haven't heard back from Sarah in 6 days."
• Call recording → action items (local Whisper, no cloud leak).
• Travel prep brief auto-compiled before any trip.
• Contact dossier before meetings.
• Family photo auto-organizer — dedupe, face-tag, NAS backup.

Tier C — Hardware (concierge, not fleet-scale)

• Home NAS + Plex — ends "iPhone storage full" forever. Home-use, 1–2 drives.
• Mac mini household AI server — local Whisper, local Ollama, no cloud leak.
• UniFi camera stack with local AI detection — home scale, not multi-building.
• 3090 box for image gen, portrait work, family art projects.

We install it, hand it over, and maintain it. We don't manage their home network, run their wifi, or troubleshoot their printer.

Tier D — Projects (one-off, higher margin)

• Internal ops platform — "BGops-for-your-business." Custom schema, multi-user web UI, role-based access, workflow forms, reports, integrations. Built from patterns we already run in production. 3–6 weeks to deploy similar.
• Custom database with web UI — replaces 6 messy spreadsheets with one real tool. Team access, forms, views, exports. 1–3 weeks.
• Ecommerce back-office — inventory + pricing + listings + competitive intel across Shopify / eBay / Amazon. Integrations + automations. 3–8 weeks.
• Websites — personal landing pages, small business, portfolio
• 3D design + print — branded giveaways, replacement parts, custom decor
• Custom Python / Flask apps — internal dashboards, workflow tools, data pipelines

Tier E — Data & intelligence (our existing stack, productized)

We already run a serious scraping + data pipeline through LobbyIQ (7.5+ GB of Canadian government, lobbying, donations, contracts, permits, regulatory data). This tier monetizes that asset and builds new scrapers on demand.

• Scrape-on-demand — need data from a site, public records, a competitor? We get it, keep it fresh, deliver it where needed.
• Data enrichment — provide a list (contacts, prospects, customers); we append everything public (LinkedIn, news, company signals, scoring).
• Competitive intelligence — monitor competitors (hiring, pricing, product, press) → monthly digest.
• Market & deal research — "tell me everything public about X company / market / person" as a standing service.
• Document intelligence — OCR + extract from contracts, leases, scanned mail; bulk summarize; searchable private archive.
• Dashboards & visualizations — raw data → a dashboard the client actually opens.
• Natural-language Q&A over their own data — RAG over emails, notes, docs, CRM.
• Gov & lobby intelligence (Canada) — tailored slices of our LobbyIQ dataset for clients who care about government contracts, political donations, regulatory filings, building permits, court filings, insider trading patterns.

05 · Architecture & privacy

Private by default — our architecture is our product

Our clientele doesn't trust "cloud AI" by default, and they shouldn't. The differentiator that justifies our pricing and hardware install fee is simple: your most sensitive data never leaves your house. We tier every piece of client data by sensitivity and route it accordingly. Under the hood it's mixed. From the client's view it's one UI with a visible indicator when anything touches the cloud.

Sensitivity tiers

The architecture this implies

Local box (Mac mini or small Linux server, ~$1,500 hardware + install)

• Whisper.cpp for call / voice transcription — raw transcripts stored on-device, encrypted at rest, never uploaded
• Ollama running Llama 3.1 70B (quantized) or Qwen 2.5 for Tier HIGH + MEDIUM reasoning
• Local vector DB (Chroma / Qdrant) for RAG over emails, notes, documents
• SSH-tunneled remote access so we can maintain without client action
• Runs 24/7, whisper-quiet, lives in their office or a closet

Cloud (Claude API) for

• Tier LOW tasks at full capability
• Tier MEDIUM after local redaction / summarization
• Heavy lifting (200k context, tool use, newest capability) when the tier permits

Hybrid flow — real example

1. Client takes a 45-min call with her accountant
2. Recorded → Whisper (local) → raw transcript stays on her box forever
3. Local Llama extracts structured data: names, dates, dollar amounts, action items
4. Client clicks "draft follow-up email" → Claude API receives the structured extract with identity tokens redacted (names become "accountant," "client"), not the raw transcript
5. Claude returns a draft in her voice, she reviews, she sends
6. Anthropic never sees the transcript

Legal reality (quick hits)

• Alberta, all Canada, UK, 39 US states = 1-party consent. Client can record their own calls without telling the other party, legally.
• 11 US states = 2-party consent (California, Florida, Washington, Maryland, Massachusetts, Montana, Nevada, New Hampshire, Pennsylvania, Illinois, Connecticut). If client talks to anyone there, consent is legally required before recording.
• Never-record list regardless of jurisdiction: lawyers, doctors, therapists, clergy, anyone delivering MNPI, any call that could waive privilege.
• Default posture: recording OFF. Client opts each relationship in ("always record my accountant," "always record client calls") or per-call. System enforces.

Operational safeguards

• Audit log — every cloud API call logged with timestamp, content hash, operation. Client reviews anytime.
• Visible indicator in UI when anything is cloud-bound. No hidden exfiltration.
• No human data review by us without explicit client approval. Debugging on their box via SSH, or on synthetic test data.
• Key management — client holds their own API keys where possible (their Anthropic account, their Google Workspace). We're the operator, not the data controller.
• NDA per client, signed before any access.
• General liability insurance covers data-handling incidents.

Why this matters for pricing & sales

• The local box is not an accessory — it's a trust artifact. The $1,500 install fee pays for itself as a trust-builder in one dinner conversation.
• Clients who understand the tiering buy more — they'll let us expand into areas they'd never give to a generic SaaS.
• The privacy story is the single biggest reason to choose us over "a guy who set up ChatGPT Team."

06 · Pricing model

Monthly retainer (core product)

Setup fees (one-time, at signing)

• Starter: $750
• Standard: $1,500
• White-Glove: $3,000

Hardware (cost + markup + install)

• Hardware at cost + 20% markup
• NAS setup: $500 install
• Home AI server: $800 install
• UniFi camera system: $1,200 install
• Full home lab (NAS + AI + cameras): $2,500+

Projects (fixed-bid or hourly)

• Internal ops platform (BGops-class): $25,000–75,000 build + $300–800/mo maintenance
• Custom database with web UI: $3,000–10,000 build + $200–400/mo maintenance
• Ecommerce back-office: $15,000–50,000 depending on platforms + SKU count + integration depth
• Personal / small-business website: $3,000–10,000
• 3D print runs: $200–500 setup + per-unit (50–100% markup)
• Custom automation beyond retainer: $150/hr or fixed bid

Tier E — Scraper pricing (three flavors)

Default: shared. Private is priced to nudge them back toward shared unless they truly need exclusivity.

Tier E — LobbyIQ data products

• Canadian gov intel brief — monthly PDF or Slack digest, custom-sliced to their industry / geography / competitors: $500–1,500/mo
• Private dashboard on {client}.lobbyiq.schlabs.dev: $3,000 build + $400/mo
• Regulatory radar — gazette + proposed legislation in their industry: $300–600/mo
• Deal sourcing feed — grants, contracts, permits filtered by geography / sector: $500–1,200/mo

Referral credit

Any existing client who refers a new signed client gets one month of retainer credited. Self-funding growth loop.

07 · Revenue sharing

Core split, per client

Rationale

• 50% delivery reflects that Curtis carries operational risk and pays execution costs (salaries, hardware capex, software subs, Claude API bills).
• 40% sales rewards pipeline. If you don't bring clients, you don't get paid — aligns incentives naturally.
• 10% reserve builds cushion before profit distributions.

Special cases

• House client (Curtis-sourced): 70% Curtis / 20% rotates between J&J for support / 10% reserve. OR 80/10/10 if no support needed.
• Joint-sourced (Justine & Josh both meaningfully involved): 25% primary / 15% supporting / 50 / 10 reserve.
• Project work: same split if rainmaker sourced; 80% Curtis / 20% reserve if Curtis-sourced.
• Referral chain: referrer's rainmaker collects the month credit; new client's rainmaker owns going forward.

Cadence & mechanics

• Monthly payouts on the 5th, after client retainers clear.
• Curtis runs payroll & 1099s. Bookkeeping in QuickBooks or Wave.
• Structure: sole-prop under Curtis until second paying client; then LLC; then Inc past $500k revenue.

LobbyIQ cross-venture flow

LobbyIQ and the consulting venture have distinct ownership (Josh holds 40% of LobbyIQ). To keep incentives clean and accounting simple, treat LobbyIQ as a licensed data vendor to consulting:

• Consulting pays LobbyIQ a flat $200–500/mo per subscribing client, depending on depth of the data slice. Booked as LobbyIQ recurring revenue.
• Consulting client pays their retainer / Tier E fee as normal; consulting margins the rest after the LobbyIQ license fee.
• Scraper IP rules:
  • Shared scrapers (public data): consulting IP. Reusable across clients.
  • Client-account scrapers: consulting IP, but only runs with that client's credentials.
  • Private dedicated scrapers: client IP. Code lives on their box.
  • Canadian-gov scrapers built for a consulting client get absorbed into LobbyIQ proper, strengthening the SaaS product. Client subsidizes LobbyIQ's feature velocity.
• Josh's dual stake (40% LobbyIQ + consulting rainmaker cut) is a clean alignment — he earns twice when he brings a gov-data-hungry client into consulting, which is exactly the incentive we want.

08 · Employee model

Year 1 — solo with contractors

Curtis does all delivery. Contractors as needed (specific 3D print runs, occasional web designer, bookkeeper for taxes).

First FT hire — around $15–20k MRR (≈12 retainer clients)

Role

Junior Automation Engineer

Profile

Python-strong, curious about Claude/AI, 2–5 yrs experience.

Comp

$85–100k CAD base + 10% bonus on retained clients (paid quarterly)

Structure

W2 (or T4). Contractor is cheaper but W2 is a trust signal for top-tier clients.

Workload

5–8 retainer clients' ongoing maintenance; Curtis focuses on new onboarding + strategy.

Second hire — around $35k MRR

Role

Hardware & on-site tech

Profile

Networking, Linux, hardware experience, strong customer-facing.

Comp

$75–95k base + 10% on install projects they run

Focus

NAS / camera / home-AI installs, on-site troubleshooting of what we installed.

Third hire — around $55k MRR

Role

Creative lead (3D + web)

Profile

Designer who codes, CAD, Figma, print production.

Comp

$75–95k + project revenue share on creative deliverables

Focus

3D design + prints, website design, creative ecommerce work.

Curtis's cut flow

From his 50% per client, Curtis covers: employee salaries, software/infra (Claude API, VPS hosting, subscriptions), tools & hardware capex, and own compensation (residual).

Once net > $20k/mo after expenses, Curtis takes formal salary ($120–180k CAD base) and retains ownership equity.

09 · Operations

Client lifecycle

1. Inbound — rainmaker identifies interest, intro call.
2. Intake — send intake.schlabs.dev URL + per-client passcode; answers land with Curtis.
3. Discovery call — 30–45 min, Curtis leads. Confirm needs, present tier options.
4. Contract signed + setup fee paid.
5. Onboarding week — initial automations, MCPs, hardware installs if applicable.
6. Proof delivered — first automation live, screen-recorded walkthrough.
7. Monthly — retainer runs. Tune-up call. New automations within tier.
8. Quarterly — strategy review with rainmaker + Curtis.

Tool stack

• Intake: intake.schlabs.dev
• Client files: private GitHub repo per client
• Project tracking: shared Notion or Linear
• Billing: Stripe recurring for retainers, Stripe invoices for one-offs
• Communication: via rainmaker, client's preferred channel
• Time tracking: Toggl (for retainer compliance only)
• Delivery platform: client-automations codebase deployed per-client on Hetzner VPS or laptop

SLAs by tier

• Starter: 72h response, monthly tune-up call
• Standard: 24h response, monthly call, on-demand advice
• White-Glove: same-day response, quarterly strategy, priority queue

10 · MRR projections

Conservative (2–3 new clients/mo once proven)

Aggressive (4–5 new clients/mo, high referral loop)

Assumptions

• Retention: 85%+ annual (concierge services retain hard if quality is consistent)
• Avg retainer grows as clients move up tiers
• One-off projects add 20–40% on top of MRR (not shown)
• Hardware installs are lumpy but high-margin — avg $5k/project, ~5/yr year one

11 · Risks & mitigations

12 · Milestones

• M0 — 2026-04-17 (today): intake.schlabs.dev live, HQ plan drafted, Wednesday session prepped.
• M1 — 2026-04-22 (Wed): Justine intake session; first automation scoped.
• M2 — 2026-05-06: Justine's first automation shipped; screen-recorded walkthrough; written case study.
• M3 — 2026-05-15: Josh + Justine begin formal outreach.
• M4 — 2026-06-01: 3 paying clients signed.
• M5 — 2026-07-15: 5 paying clients, ~$3–4k MRR.
• M6 — 2026-10-01: Entity formed. Stripe recurring live. GL insurance in place.
• M7 — 2027-04-01: $15–20k MRR. First engineer hired.
• M8 — 2027-10-01: $30k+ MRR. $200k+ annual run-rate.

13 · Open questions

Decisions for the three of us before we sign client #2. Leave comments under each section as you think through these.

1. Entity structure — LLC, Canadian Inc, or partnership? Single-owner Curtis with profit-share to J&J, or formal 3-way partnership with ownership?
2. Firm name — do we brand, or ride on personal names until it emerges? Options: Lantern, Atlas, Meridian, Signal, Schlabs Concierge, or something else.
3. Territory / attribution rules — first-touch-wins the cleanest default; 90-day reset if no progress.
4. Justine's month 1 — free pilot, or standard pricing? Recommendation: free 30-day pilot, standard after.
5. Branding visibility — clients see a 3-person firm, or only "their person"? Recommendation: single-face per client, Curtis visible on technical calls.
6. Exclusivity — do J&J commit to this as priority, or parallel other consulting / sales?
7. Equity vs revenue share — 50/40/10 per-client is revenue. Equity stakes (if we ever sell) separate? Recommendation: align equity to revenue share (Curtis 50%, J&J 20% each, 10% option pool).
8. Marketing posture — visible firm with website & case studies, or strict word-of-mouth? Our clientele values discretion, so "barely a website, referral only" is viable and possibly more prestigious.
9. LobbyIQ licensing model — flat per-client fee (simple, predictable) vs. revenue share (rewards bigger clients but complicates accounting). Recommendation: flat fee at launch, revisit at 10+ clients.
10. Private scraper IP — always fully the client's (current default), or licensed back to us with a royalty if the target becomes broadly valuable? Royalty model unlocks upside but clients may balk.