danielslater.dev — Business Model & 12-Month Roadmap

Produced 2026-07-17 by a multi-agent workflow: 5 Sonnet research legs (pricing, packaging, customers, acquisition, run-costs) → 3 Opus strategy designs (cash-first / recurring-first / leverage-first) → 3 Opus adversarial critiques → 1 Opus synthesis. ~620k tokens. Raw research data: business-model-research-2026-07-17.json next to this file.

Daniel's interview decisions (inputs, settled 2026-07-17)

• Goal: side income first — R20–50k/mo within 12 months; day job stays.
• Capacity: 10–20 h/week total (sales + delivery + support + everything).
• Markets: both — SA/Cape Town base + UK/EU/US remote margin.
• ⛔ Hard exclusion: NO hospitality clients (employer separation). Restaurant guest app = capability proof only.
• Channels: willing to run all four (warm network, agency white-label, content, outbound/platforms) — plan sequences them.
• Recurring: managed services are the destination, but only after the client- grade substrate (monitoring, metering, billing) is built; interim = best-effort retainers, no hard SLAs.
• Stack: self-host from day one (own repos/patterns only — get written OK before reusing anything Nsuna-owned).
• Names cleared: Berlein Psychology, Aerias, Nsuna all usable publicly.
• Domain email: daniel@danielslater.dev being set up this week (COMMANDS.md ▶3).
• Pipeline today: 1–2 warm conversations; Alexa + Aerias referral asks possible this month.
• Tutoring: out of scope (route already just redirects home).

RECOMMENDED BUSINESS MODEL — danielslater.dev

Backbone chosen: leverage-first (web-Repaint lead → referral-then-white-label channel → chat-over-voice AI → concurrency cap). Grafted in: the paid Signal Audit and recurring-native framing from recurring-first; the FX/platform-take realism and voice-sequencing discipline from cash-first. Dropped (killed by critiques): delivery-only effective rates, the "one INTL project clears R50k/mo" claim, 100% retainer attach, the Designjoy turnaround promise, voice-as-lead, and the restaurant-referral seed. All hours below are all-in (build + revisions + discovery/SOW/invoicing/handoff + support + substrate), not delivery-only — the single error all three candidates made.

Honest achievable band up front: the corrected arithmetic (adopted from the leverage-first and recurring-first critiques) says the sustainable landing zone is ~R25–35k/mo net by month 12 at 14–17 h/wk, R20k reachable by ~month 6. R40–50k/mo is a peak-deal-month ceiling, not a monthly baseline — it is reached by raising the INTL price anchor once references exist, not by adding volume, because the binding constraint is qualified deal flow × the 20 h/wk ceiling, not price-per-job. The brief's R20–50k band is met at its floor-to-middle sustainably; its top only intermittently.

POSITIONING & LEAD OFFER

Lead offer: Websites & web apps, led by the "Repaint" (rebuild/modernize an existing site or web app on a modern stack). AI chat agents are the recurring-native second offer; ops/automation is opportunistic third; mobile is a MobiLoud-style wrapper, on-request only; voice is deferred to Q3 and capped (see below).

Why Repaint leads (research-ranked, all four critiques endorsed this call):

• Highest-confidence arbitrage of the four (pricing leg, verbatim): a $2,750–4,500 quote to a UK/US client lands 40–70% below their local agency floor (US agency $10k–35k, UK £3k–12k; US freelancer $2k–8k) while earning 3–8× the equivalent SA-market job (SA basic site R12–40k). AI and automation have no SA-specific pricing in the research — a genuine gap — so their arbitrage is unproven; web's is documented.
• One-shot deliverable → low support tail (packaging leg: web is "naturally one-shot"; AI/automation carry the open-ended "it broke" tickets). This is what protects the day job.
• Universal white-label handoff: agencies subcontract web build capacity, not AI/mobile — so it feeds the durable channel (below).
• Two live in-category references already (berleinpsych.com, aerias.co.za) → fastest referral time-to-cash (days–weeks).

Why chat, not voice, carries the AI offer: voice's strongest documented trigger is restaurant missed-calls (~$15,750/mo) — the hard-excluded vertical — and the other universal missed-call vertical (trades) is commoditized at $29–349/mo. Chat agents are near-100% margin (tokens <$5/mo at 1,000 conversations), async (genuinely "best-effort, no SLA" honorable next to a full-time job), and need a far lighter substrate. Voice carries an implicit real-time SLA a full-time employee cannot honor and needs metering/monitoring that doesn't exist until Q3 — so it is deferred and only built if paid demand appears.

One-sentence positioning (stack-agnostic, outcome language): "I rebuild the websites and web apps businesses already run on — faster and cheaper than a from-scratch rebuild — and wire in the AI that answers their customers and recovers their lost sales, then keep it running for a flat monthly fee."

ICP

Three profiles. Restaurants/hospitality hard-excluded across all three; the restaurant app is anonymized capability proof only (never a referral seed — see Risk 5).

ICP-A — SA / Cape Town health, wellness & professional-services SMB (ignition + references). Vertical × geo: psychology/dental/allied-health practices, small B2B/advisory, local retail — SA local, ZAR. Trigger: dated/mobile-broken site, lost their original developer, or a rebrand forces a redo; and/or the "tested-but-not-embedded" AI gap (dental 47% tested, <11% embedded). Owner-operator decision-maker; #2 objection is ROI-uncertainty (24%), answered with Daniel's own before/after numbers. Reference: berleinpsych.com. This is the Repaint + Care Plan + Chat buyer and the fastest cash (warm network, days–weeks). Legal/accounting deliberately excluded — 35–41% already served by compliance-aware specialists, weak generalist wedge.

ICP-B — UK/EU/US e-commerce SMB (arbitrage + recurring). Vertical × geo: Shopify/Woo owner-operators, remote. Trigger: cart abandonment + rising support-ticket volume handled after-hours. Grounded pain (best geography-agnostic wedge): 83–84% chatbot adoption, $3.50 return per $1, 10–15% cart recovery, ~$24k/yr SMB savings. Reference: aerias.co.za. This is the Repaint/Storefront + Chat (cart-recovery) buyer — where arbitrage and the recurring attach compound. Reached via white-label agency (Q2) then Upwork (Q3), not chased cold early.

ICP-C — UK/EU boutique design/marketing agency, 5–25 people (durable engine). Vertical × geo: the agency itself, UK/EU (better day rates than SA). Trigger: won a build/rebuild they can't staff, or their current dev subcontractor is too expensive/flaked, and they need white-label overflow to hit a deadline without a permanent hire. Marginal CAC ≈ 0 after the partnership is live (acquisition leg); cost is the 4–12-week landing. Daniel owns the recurring/hosting layer as the anti-capture moat.

PRODUCT LADDER

Fixed packages, published scope, not hourly (research: AI/automation price as fixed build+retainer; published scope removes the per-lead discovery call that eats the hour budget). Internal shadow rate R900/hr for margin sanity-checks and out-of-scope overflow only — never quoted. Change-order rate for anything off-scope: R900/hr SA · $110/hr INTL. INTL prices are the invoice currency; ZAR shown at spot R16.33/$ (£1 = $1.347 = R22.00), but planning/realized math uses an R16.0/$ buffer against rand strengthening (2026 range R15.73–17.19).

Agency wholesale card (ICP-C): all build packages at retail − 25–30% (Repaint → R20,000–21,000 / ~$1,950), plus a bespoke day rate R7,000 / $430. Daniel keeps delivery margin, spends ~0 sales hours, and retains the hosting/retainer layer (the stickiest thing the agency can't cheaply replace). Note: research grounds only the ~10% pure-referral commission norm; the 25–30% true-white-label discount is a market-plausible estimate — flagged.

Pricing grounding: A/B/C sit inside the SA guides (basic R12–40k, WooCommerce R40–120k) and below the UK/US floor. D/E follow the market build+retainer pattern (custom single-purpose agent $1,500–5,000 build + $300–800/mo — priced at the low end for part-time scope and easy close). No SA-specific AI or automation pricing exists in the research; D/E/F SA prices are set off the SA web-dev hourly proxy (R450–950/hr) and stated as such. F mirrors Make.com $2,500–3,500. G anchors on SA maintenance R500–1,500/mo and UK £50–300/mo. H undercuts MobiLoud ($399–799/mo + $850–1,500 setup).

CHANNEL PLAN

Sequenced, not parallel-four. Referral = ignition (fast cash + references); white-label = durable engine (near-0 marginal CAC); Upwork = the INTL leg (deferred past its cold-start tax). Cold email and BNI dropped with the numbers. Only referral + agency have close rates high enough for the hours ceiling (acquisition leg).

Dropped, with evidence: Cold email — 3–5% reply, 0.215% end-to-end close (~1 deal / 464 emails), 30–70+ h/client. BNI — R20,300–26,900/yr cash + 75–100 h/yr, 3–6 mo to trust. Both fatal against the hours ceiling and near-zero-spend preference.

Never run more than TWO channels hot at once — sales budget is ~4–6 h of the 15 h/wk. Hospitality decline stated in every agency-partner brief and the intake form. SA leads first (ignition, ZAR ballast); INTL is the margin layer that ramps Q3.

UNIT ECONOMICS

All-in Daniel-hours (build + revisions + discovery/SOW/invoicing/handoff), the correction the critiques demanded. Effective rates are net of these hours; INTL rows show realized rates net of channel take (Upwork 15% + ~2% Payoneer + R16.0/$ buffer; agency wholesale −27.5%). Infra grounded in the costs leg. Research has no hours-per-client data — every hour figure is a measurement target for clients 1–3, deliberately estimated conservatively (higher than the candidates used).

Recurring is the floor/ballast, NOT the rate leader (correcting recurring-first's "near-zero-effort" framing, which the critique showed contradicts the time-scarcity premise). At a realistic 1.75 h/mo/client (not the candidates' 1 h — no data exists, so budget conservatively), a Care Plan nets R771/hr — below the project rate. Recurring buys revenue smoothing against project lumpiness, not a superior hourly.

Monthly math — honest, all-in hours, net of take. Planning baseline 15 h/wk = 65 h/mo, leaving 5 h/wk slack under the 20 h ceiling for day-job spikes and overruns.

The steady scenario lands ~R35k net at ~11 h/wk — consistent with the leverage-first critique's independently-derived "~R37k at ~50% overhead." The binding constraint is deal flow, not hours: 0.8 Repaints/mo (1 every ~5 weeks) sustained is the risk, and H1 will have dry months (recurring floor ~R5k) until the book builds. Recurring floors at ~R9–14k net by Q4 — covering roughly half the R20k floor before any project closes (correcting recurring-first's R19,900 gross claim: recurring alone does not clear R20k). Pushing toward R50k/mo sustainably means raising the INTL Repaint anchor to $3,500–4,500 once references exist — a price lever, not a volume lever (volume is hours-capped).

12-MONTH ROADMAP

Daniel already runs 24/7 personal agent infra (VPS, systemd, Discord bot, event log, SQLite) — the monitoring/alerting muscle is genuinely de-risked (Q2 reuses his event-log/Discord pattern). The client-grade substrate is ~80–110 h total, staged Q1–Q3, and IS budgeted in the hours above (correcting all three candidates, who costed it at zero). Leading with chat + web (not voice) is what keeps the substrate small and off the critical path — the voice substrate is built only if paid voice demand appears (Q3+).

RISKS

1. Deal flow is the binding constraint and revenue is lumpy (revenue resets monthly; H1 has dry months). Mitigation: build the R9–14k/mo recurring floor to cover half the R20k target; keep referral + one agency partner always warm; maintain a standing 3-deal pipeline; push toward R50k via INTL price anchor, not volume (volume is hours-capped).
2. Fully-loaded hours breach the 20 h ceiling in peak-deal months (all three candidates hid this; peak scenario = ~19 h/wk). Mitigation: concurrency cap = 1 active build + waitlist (no Designjoy turnaround promise — Daniel is ¼–½ a full-timer's throughput); Storefront is "occasional big fish"; substrate front-loaded to Q1–Q2 so steady-state substrate ≈ 2 h/mo; the 5 h/wk slack absorbs overruns.
3. Scope creep reverts productized offers to bespoke within ~90 days (research #1 failure mode; $7,800–15,600/yr lost); worse through agencies, who own the client and negotiate scope you can't police. Mitigation: published fixed scope + hard change-order rate (R900/h · $110/h) + paid Signal Audit that writes scope down first; "upsell to the next package, never bend the current one"; in agency contracts, scope changes route back to Daniel for re-quote in writing.
4. Managed-service / SLA obligation collides with the full-time job (real-time voice cannot be triaged during a day-job standup; retainers sold before monitoring exists). Mitigation: lead with chat + web maintenance (async, honestly best-effort), defer voice to Q3 and only after Q2 monitoring + Q3 metering ship; no hard SLA until Q4; minute caps + overage on any voice; status page is Q3, not a Q1 promise.
5. Hospitality/employer-adjacency leak through referral + agency channels (research's #1 vertical is restaurants; Daniel's employer is hospitality-tech; referral/agency channels don't vertical-filter). Mitigation: restaurant app = anonymized capability proof ONLY, never a referral seed — do NOT ask restaurant contacts for intros (their network is hospitality = the exclusion + an employer conflict of interest); referral seeds are berlein + aerias only; hospitality decline written into intake + every agency brief; confirm the restaurant app is Daniel's own IP (not built on employer time/assets) before any portfolio use.

FX note (not a top-5 risk): rand ranged R15.73–17.19 in 2026 (spot R16.33). Invoice INTL in USD/GBP, plan at an R16.0/$ buffer, take 50% upfront, use Wise/Payoneer (~2%); ZAR clients are the stable ballast. R20–50k = ~$1,225–3,060 at spot — ~10% higher in dollar terms than the R18/$ framing implied.

KPI CHECKPOINTS

DECISIONS FOR DANIEL

1. Ignition market: SA-first (recommended) vs INTL-first. Rec: SA warm network for cash + references in Q1–Q2 (days–weeks to first client, ZAR ballast), INTL as the margin layer ramping Q3. If INTL-first instead: higher R/h sooner but a 15–30 h Upwork cold-start with no reviews and dry H1 — only choose if the SA network genuinely yields nothing by M1 (this is the M1 pivot).
2. Lead offer confirm: Repaint-lead + chat-attach (recommended) vs recurring-native AI-lead. Rec: Repaint leads (documented arbitrage, low support tail, live references); chat is the recurring attach. If AI-lead instead: better recurring-native economics on paper but front-loads the substrate and the SLA collision — the critiques showed this pushes fully-loaded hours to 23–30 h/wk. Only revisit after the substrate exists (Q4+).
3. Voice: build the Q3 voice substrate or stay chat-only. Rec: stay chat-only unless ≥2 paying voice leads appear — then build metering+monitoring first. If you build voice speculatively: +30 h substrate and a real-time SLA obligation against a full-time job, for an offer whose best trigger (restaurants) is excluded.
4. White-label depth: 25–30% true-white-label wholesale (recommended) vs 10% referral-only. Rec: offer both — 10% finder's fee when you keep the client relationship, 25–30% when the agency fully resells under their brand — and always keep the hosting/retainer layer as the anti-capture moat. If referral-only: higher margin per deal but the agency has less incentive to feed you volume. (The 25–30% figure is estimated, not in the research — validate against what partners actually accept.)
5. INTL price anchor: hold $2,750 Repaint (recommended for first 3–5 INTL clients) vs open at $3,500–4,500. Rec: start low to win the first INTL references fast, then raise — this is the lever toward the R50k top of the band. If you open high: better margin but slower first-INTL close and less review-compounding on Upwork.
6. Run the follow-up SA AI/automation pricing search before publishing D/E/F SA prices. Rec: yes — the research has no SA-specific data for these; current SA prices are a web-dev-hourly proxy. If you skip it: you risk mispricing the two offers with the thinnest grounding. (This is also flagged in the research's own questions-for-Daniel.)
7. Confirm restaurant-app IP provenance before any portfolio use. Rec: verify it wasn't built on employer time/IP; if clean, use anonymized as capability proof only. If it's employer-entangled: drop it from the portfolio entirely and demo voice capability with a purpose-built sample instead — the employer conflict outweighs the proof value.

ADDENDUM 2026-07-17 (late) — WhatsApp as the SA chat surface

Daniel's hypothesis ("WhatsApp ordering/booking/AI agents would be a hit in SA") was researched the same evening (2 legs, research/2026-07-17-whatsapp-sa.md). Verdict: HOLDS — WhatsApp becomes the delivery surface for the SA chat-agent offer, with the health-practice booking agent as the flagship. Re-scoped as follows.

Why it holds

• WhatsApp is SA's default channel, not an option: 96% of SA internet users, 28M users, users' single favourite app (1 in 3), ~23h42m/month each. A web-chat widget is the wrong surface for ICP-A; WhatsApp is where their patients already are.
• The exact gap matches ICP-A: SA health WhatsApp solutions today are aggregators (RecoMed, WhatsUpDoctor) — no affordable "WhatsApp receptionist wired to YOUR calendar" product exists for individual practices. Generic SaaS bots sit at R299–2,000/mo (BizAI, Karabo.ai, GotBot…); bespoke agency builds start ~R180k. Daniel's R15k + retainer sits in the empty middle — sold on done-for-you custom integration + accountability, never on "has a WhatsApp bot".
• Meta policy blesses this exact use: since Jan 2026, general-purpose bots are banned but narrow booking/ordering/support bots are explicitly allowed — the bot must be architecturally scoped to refuse off-topic requests (compliance, not UX).
• Proof point for sales: WhatsApp booking/FAQ bots cut front-desk call volume 40–60% in professional services (bizai.co.za) — an ROI number for the pitch.

The two clocks (both material, both this quarter)

1. Oct 1, 2026 — Meta meters the free service window. All the competitor pricing and blog math assumes free in-window replies. Price on post-October economics from day one or margin quietly evaporates on live clients in Q4.
2. Aug 1, 2026 — Meta's native "Business Agent" starts billing ($2/1M tokens, ~4–5c/message) — Meta itself commoditizes the generic bot. The pitch must lead with what native can't do: real calendar/PM-system integration, SA payment provider of choice, CRM push, POPIA-aware health-adjacent handling, bespoke flows.

Re-cut pricing (post-Oct-2026 cost regime, lean stack)

Stack: self-built orchestration (Daniel's existing agent-infra pattern) + pay-as-you-go BSP (Twilio ~$0.005/msg markup, no monthly) for low-volume clients — NOT a bundled platform (WATI alone ≈ R1,400/mo, eats the whole retainer). 360dialog (€49/mo flat, zero markup) only wins at higher volumes.

• Chat Agent (D) SA price becomes: R15,000 build + R1,800/mo "WhatsApp Receptionist" (transactional/booking scope ONLY — broadcasts/marketing messages are a separately priced line item at Meta-cost + margin, since marketing runs ~R1.40/msg and 5k msgs/mo ≈ R7.5k in Meta fees alone).
• Web-chat variant stays R1,500/mo for intl e-comm (no Meta/BSP fees, near-pure margin).

Build/sales realities to bake into scope

• Onboarding lead time: Meta Business verification 2–14 days + template review + a dedicated number (client can't reuse their personal WhatsApp number) — set expectations pre-signature; no same-day go-live.
• Payments: no native WhatsApp Pay in SA — flows use hosted payment links (Yoco ~2.6% / Paystack 2.9%+R1.50 / PayFast) + webhook confirmation back in chat. Demo it honestly as "tap, pay, get confirmed here", never "in-chat checkout".
• POPIA is hygiene, not a differentiator (incumbents ship it) — implement real opt-in, minimisation, deletion rights, human-in-the-loop for health-adjacent calls.
• Open gap: what scheduling/PM software SA practices actually run (Google Calendar vs closed practice-management systems) — determines integration effort; needs either a research leg or answers from the first 3 Signal Audits.

Reference & proof engine (Daniel: "wish we had more references")

1. Case-study one-pagers for berlein + aerias with before/after numbers (Q1, already in roadmap).
2. One discounted "founding client" per package, traded explicitly for a named case study + metrics rights — cheaper than any other CAC and fixes the proof gap.
3. A live WhatsApp Receptionist demo number for a fictional practice — doubles as the site's AI-section demo and the strongest possible sales asset ("message it now").
4. Signal Audits produce before/after numbers even when the build doesn't close — every audit feeds the proof library.

Integration reality (practice-tools leg, same evening — research/2026-07-17-sa-practice-tools.md)

The WhatsApp Receptionist pitch splits by sub-vertical — one product, two stories:

• Competitive check: Bookem (SA, PM + AI-drafted notes + automated booking messaging) overlaps most — scope it before the first dental pitch. BizAI Voice Valet (R999/mo AI voice, 24/7) is the cheap-AI floor. R1,800/mo also exceeds many practices' ENTIRE PM subscription — the premium is carried by WhatsApp-native + real calendar write-back + done-for-you accountability, never "has a bot".
• Willingness-to-pay caveat: the receptionist-salary comparison only holds where a receptionist exists (dentists). For psychologists, test R1,800/mo against real prospects (the first Signal Audits — and Alexa is exactly this segment) before treating it as validated.
• Demo-build directive: the live demo (next build task) should use Google Calendar sync as its integration — the psychologist-segment MVP path, and the demo persona is a psychology practice anyway.