01The frame — working, not final
What's been provisionally agreed, and is still open to challenge.
Lead with the SMB's outcome, in their language. AI is the engine, never the message.
This was arrived at adversarially — the operator's lens (AI is the unfair advantage) argued against the buyer's lens (the buyer wants the result, not the technology), and the synthesis held: AI is the proof, never the promise. Two doors — outcomes to buyers, AI to peers — never mixed. That frame is the floor everything below is built on. It is the current answer, not a permanent one.
02The room — the lenses arguing
No single voice gets to decide.
Connie · analyst
Baseline-first, MECE, no directional answers. Lays out the option space and the evidence; refuses to pick for you.
Rick · the buyer's adversary
"They don't care about AI — they want more customers." Strips the operator's romance off every claim. Demands outcomes.
The operator
AI is the edge that lets one person deliver agency-grade work at 4 days/week. Argues the moat, not the message.
Kendall · the teardown
Takes the chosen answer and tries to break it. The verdict only counts once it has survived this.
The process is the argument between them: steelman both sides, derive rather than assert, then test against reality. A position that hasn't survived Rick and Kendall is not yet an answer.
03The method — how an answer earns its place
Four disciplines, applied to every candidate.
- Derive, don't pick. A wedge is scored against criteria — speed-to-visible-cash, clean attribution, low trust-barrier, self-funding, defensible vs the commodity tools, fit to the craft and the 4-day week, path to recurring revenue. The winner emerges from the table; it isn't announced.
- Let proof-state choose the offer. The entry mechanic (free pilot → free diagnosis → paid+guarantee → rev-share) isn't a preference — it's selected by how much proof exists and how scarce the operator's time is.
- Measured, not reported. "Successful" is a forecast until a number is in. Nothing counts as proof until it's measured.
- Everything is a bet, stated as a bet. Each claim carries its own falsification — the thing that, if true, kills it.
The market backdrop the whole thing sits in: the price line is collapsing (AI-agent "agency replacement" at $99–499/mo), AI eats execution while trust, taste and distribution stay scarce, and Dubai is a named AI-marketing hub in a growing region. Sourced in the deep dives.
04The segments — steelmanned, not ranked
Each candidate gets its strongest case and its sharpest counter. No crown.
Aesthetic / cosmetic clinics
Tier 1 · candidateStatus: leading on economics, unproven on access. The counter is winning until a warm door appears.
The worked argument →Dubai real estate (brokerages + boutique developers)
Tier 1 · candidateStatus: strongest on access and proof-in-flight; weakest on whether the proof transfers and banks. Both live.
The worked argument →Boutique off-plan developers
Tier 1 · candidateStatus: an expansion segment, not a beachhead. Park behind a proven motion.
Tier-2 watchlist
laterEducation/training (bilingual, seasonal, slow) · interior design & fit-out (project-based, smaller volume) · hospitality groups (booking-driven, thinner margins). Real, but none beat the Tier-1 candidates on speed-to-proof. Held, not pursued.
Deliberately excluded — naming what's OUT is process too
Tier 3 · avoidRegulated (banking, insurance, legal — procurement kills the SMB offer) · generic e-commerce (agency price war) · retail (margins too thin) · anyone with a procurement department. Excluding these on purpose is part of the discipline, not an omission.
05The open bets — the live arguments
The cross-cutting tensions that aren't resolved.
STILL BEING ARGUED:
- Opener vs moat. The fastest proof (reactivation) and the most defensible thing (editorial brand) are different outcomes. Open on one, expand into the other — but which leads?
- Access vs economics. Clinics win on the spreadsheet; real estate wins on the warm door. Access usually beats theory — does it here?
- Recurring vs project. The sticky money is a retainer; the proven muscle is project work. Which business does the operator actually want to be?
- Proof transfer. Does a luxury case study convert a mid-market buyer, or does it read as "not for me"?
- The free-consult trap. AI Sherpa is the most on-moat funnel and the easiest way to give work away. The close has to be structured or the motion bleeds time.
- Identity vs strategy. "AI-marketing consultancy" is the operator's own brand — and the strategy says hide AI from buyers. The two doors reconcile it only if the public surfaces get re-pointed.
06Where the process itself could fail
Kendall, turned on the method.
The biggest risk isn't in any single segment — it's the pattern. Building artefacts instead of selling. This hub, the two deep dives, the research digest — all genuinely useful, and all at risk of being the thing that gets done instead of a conversation with a buyer. A process this clean can become a very well-argued way of not picking up the phone. The method is only worth anything if it ends in contact with a real customer this week.
07What actually decides it
Reality, not argument.
Three tests collapse most of the uncertainty above, and none of them is another document:
- Find one warm clinic door — or concede the segment to its counter and commit to real estate.
- Bank one BossCouple number — a measured reactivation figure or response-time lift turns the strongest forecast into the strongest fact.
- Run the first five AI Sherpa conversations — each closing on a free diagnosis — and see if the engine converts before scaling to 50.
The hub holds the thinking so the thinking stops being the work. The next move is a buyer, not a better page.