How I AIHow a VC and tech founder used AI to launch a brick-and-mortar business in their spare time
CHAPTERS
- 0:00 – 6:36
Board-game renaissance and the idea for a membership-based tabletop social club
Andrew and Nabeel describe how modern board games became a surprisingly social, mainstream hobby—and why organizing game nights is harder than it should be. They land on the concept of a physical, membership-based club where people can drop in, bring friends, or be matched with other members.
- •Board games as an off-screen social activity with growing popularity
- •Pain point: coordinating game nights amid busy schedules
- •Concept: a physical membership club vs. a typical board game café
- •Key differentiator: help members find others to play with on demand
- 6:36 – 8:07
Why AI made the “ridiculous” side project feasible at all
They explain that the business likely wouldn’t exist without AI because neither founder had small-business experience or the time for a full second job. Claude (and other tools) lowered the activation energy by turning vague ideas into actionable plans and doing large volumes of work quickly.
- •AI as the catalyst to go from dinner idea to signed lease
- •Claude as an encouraging, practical “you can do that” partner
- •Time constraints: families + full-time jobs make manual execution unrealistic
- •AI makes it possible to indulge ambitious ideas with faster iteration
- 8:07 – 10:20
Claude Projects as a business copilot: documents, artifacts, and institutional memory
Nabeel outlines a workflow centered on Claude Projects: generate documents, then reuse them as context for subsequent decisions. Instead of one perfect prompt, they iteratively build mission, plan, and strategy artifacts that become the evolving source of truth.
- •Three core AI roles: copilot planner, grunt worker, experience-enabler
- •Preference for document generation over long conversational threads
- •Iterative loop: create artifact → add to project context → refine next step
- •Fast feedback cycles enable sustained creative momentum
- 10:20 – 13:03
The practical planning backlog: budgets, timelines, LOIs, decks, and operational checklists
A tour through the kinds of real-world details that appear when opening a brick-and-mortar business. Claude helps draft landlord communications, assemble timelines, and produce materials that make the venture look credible—without requiring prior retail expertise.
- •Space layout and budget thinking (with caveats)
- •Drafting a letter of intent to a landlord
- •Permitting timelines and operational to-do lists
- •Creating a landlord-facing pitch deck and other business collateral
- 13:03 – 13:20
Where AI breaks down: floor plans and the limits of spatial reasoning
They candidly discuss what didn’t work well—especially floor-planning—because LLMs lack true spatial understanding. They emphasize the importance of learning tool boundaries by testing multiple models/products and adapting expectations.
- •LLMs struggle with spatial memory and accurate floor plans
- •They experimented across tools (Claude, Cursor/Windsurf, others)
- •Value comes from knowing edges and switching approaches
- •Using AI where it has internet-learned patterns (planning/personas) works best
- 13:20 – 16:39
Customer personas and programming events to serve different player types
Claude helps them build a persona matrix based on game variety preference and introversion/extroversion. They use it to design events and offerings beyond the typical “Magic night” model, aiming for a broader mainstream audience.
- •Persona axes: devoted-to-one-game vs. novelty-seeking; introvert vs. extrovert
- •Estimating likely customer mix for a board-game social club
- •Designing event formats to match different social comfort levels
- •Avoiding the narrow ‘one hardcore community’ tuning of many game shops
- 16:39 – 21:02
Checking viability: capacity, utilization, and ‘not a money pit’ economics
Before (and after) committing, they use AI to reason about whether the club can break even. AI helps them figure out core drivers like how many people can fit, what utilization might look like, and what pricing and programming could support sustainability.
- •Goal: break-even, not a unicorn outcome
- •Capacity constraints: how many people/tables can the space support?
- •Using AI for pricing frameworks and rough financial projections
- •AI shifts mindset from ‘cool idea’ to testable business hypothesis
- 21:02 – 25:53
Berkeley real estate, legalities, and permitting: using AI for confidence and second opinions
They still hire experts (lawyers, agents), but use AI to interpret ordinances, sanity-check requirements, and reduce fear of getting blindsided. AI functions like a ‘second opinion’ that makes complex local bureaucracy navigable enough to proceed.
- •AI complements—not replaces—lawyers and real estate professionals
- •Copy/pasting ordinance pages into Claude for interpretation
- •Permitting category questions: what approvals will take months vs. years?
- •Main benefit: confidence to start and keep momentum in a Byzantine process
- 25:53 – 28:24
From SaaS shopping to building: deciding to create an AI concierge instead of buying tools
They first try to find existing reservation/co-working/table-management software, but nothing fits their custom ‘find me players’ vision. They realize the system is essentially a small database plus a conversational interface, making a bespoke build realistic.
- •Initial approach: evaluate off-the-shelf reservation and management SaaS
- •Gap: booking a table is easy; matchmaking/LFG workflows are not
- •Reframe: simple tables + simple DB + chatbot layer
- •AI makes custom software viable for a side project
- 28:24 – 32:30
Architecture overview: Airtable as the database, n8n as the workflow engine
Andrew demos the backend: Airtable stores members, games, tables, and reservations, while n8n orchestrates the agent workflow. They highlight why Airtable is practical for non-engineers and staff, and how AI reduces the pain of ‘messy’ fields like availability and preferences.
- •Airtable tables: users/members, games inventory, physical tables, reservations
- •Free-text fields become workable because AI can interpret them
- •Why Airtable over Postgres: fast GUI, human-operable admin interfaces
- •n8n enables integrations without building a full custom frontend
- 32:30 – 34:10
Populating and merchandising the game library: AI categories and a curated retail experience
They use AI to create playful, bookstore-style merchandising categories and to assign games into mutually exclusive groupings. This kind of detailed curation is usually abandoned due to effort, but becomes feasible with AI-assisted tagging and automation.
- •Curated category concepts (e.g., mood/experience-based shelves)
- •Avoiding duplicates by making categories mutually exclusive
- •Using AI to suggest which games fit each category
- •Airtable’s AI features + Claude help make this ‘worth doing’
- 34:10 – 36:21
A Dewey Decimal System for board games: generating TLCS codes at scale
Nabeel explains their “Tabletop Library Classification System,” assigning each game a structured code that maps to mechanics and complexity. AI enables categorizing hundreds of games consistently—something they argue would be impractical manually.
- •Library analogy: help newcomers navigate ‘a wall of games’
- •TLCS structure: major category (e.g., Cooperative) → subcategory → complexity weight
- •Example: 400s cooperative; 420 adventure co-op; .5 weight/complexity
- •AI enables consistent classification across large catalogs
- 36:21 – 39:47
Demo: texting the AI concierge to schedule games and matchmake with members
They walk through the customer experience: text a request in natural language, and the agent creates a session, books a table, and recruits compatible members. The flow uses Twilio for SMS, n8n for orchestration, and Airtable (via MCP tools) as the system of record.
- •User sends SMS request (time window + game type)
- •Agent creates a request and reaches out to matching members
- •Quorum-based reveal: if enough players commit, the session is confirmed
- •Tooling: Twilio intake → n8n routing → Airtable MCP create/list records
- 39:47 – 48:54
What AI enables beyond automation: new experiences, minimal code, and tips for better prompting
They reflect on AI’s impact: it’s not about replacing a person, but enabling coordination and curation that wouldn’t exist otherwise. They close with favorites, and practical advice for when models ‘don’t listen,’ emphasizing context-building and prompt iteration.
- •AI enables services that are too slow/expensive for humans to run manually
- •Most progress comes from prompts + tool wiring; some code still exists (Python, n8n)
- •Viability tooling includes more advanced modeling (capacity planning, optimization)
- •Prompting tips: reset when stuck; add context/‘throw tokens’ and iterate
