Build a Local Discovery Platform: The near.now Playbook

The problem

Google Maps owned local search for two decades. Over 80% of consumers used it to find nearby businesses. The position seemed permanent. Then AI agents arrived and changed the question entirely.

When a user says “find me a plumber this afternoon,” they are no longer typing a query into a search box. They are issuing an instruction to an AI assistant. The assistant does not return ten blue links. It does not show a map with twenty pins. It selects one provider, justifies the choice, and initiates the booking. The user never sees a search engine results page. Over 65% of local searches in 2026 now happen via voice-activated queries — and the AI assistant reading the result aloud means “Position Zero” (the single best answer) is the only position that matters.

The Location-Based Services market projects $150B+ by 2033, growing at 19.6% CAGR. The local discovery layer that AI agents trust will capture the transaction value; the platform that AI agents consult first wins. Google Maps was built to answer “what is nearby.” The next-generation platform answers “who should I hire.”

This walks through the architecture of a platform that AI agents recommend instead of Google Maps — and how subscription revenue, per-booking fees, and sponsored listings combine into a $141K/month conservative model.

What most people get wrong

Mistake one: competing with Google Maps on mapping data. Local discovery startups historically tried to out-map Google — better business listings, more reviews, finer-grained category trees. Google has spent billions on mapping data; the gap is uncatchable. Competing on the data dimension Google has already won is a losing game.

The fix is to compete on a different dimension: machine-readability for AI agents. A human can interpret a messy website, overlook inconsistent hours, and forgive an outdated menu. An AI agent treats inconsistent data as a risk signal and deprioritizes it. The five requirements for an AI-agent-ready business — structured business data, real-time availability, cryptographically verified identity, contextual metadata, and verified trust signals — are exactly the dimensions Google Maps neglects. A platform that delivers these to AI agents becomes their default local source.

Mistake two: starting with breadth instead of density. The instinct on launch is to cover ten cities at once because “national” feels like the proper scope of a local discovery platform. The pattern fails: each city has too few providers to satisfy demand, AI agents recommend providers from out-of-area, customer experience degrades, and the brand dies under the weight of empty markets.

The fix is density before breadth. Pick one city. Recruit the first 100 providers across the highest-demand categories. Hit a service-density threshold (typically 100 providers across 5 categories) before opening Market 2. The book’s launch playbook lays this out city-by-city; the conservative model reaches $141K/month with three cities operating at density.

A working approach

The five requirements for machine-readability are the platform’s value-add to providers. A business profile that meets them looks like this:

interface LocalBusinessProfile {
  // Identity (cryptographically verified)
  did: string;              // did:tp:near.now:provider:abc123
  name: string;
  category: ServiceCategory[];
  verifiedSince: Date;

  // Location Intelligence
  location: {
    lat: number;
    lng: number;
    serviceRadius: number;  // km
    zones: string[];        // "downtown", "east-side"
  };

  // Real-Time State
  availability: {
    nextSlot: Date;
    slotsToday: number;
    acceptingNewClients: boolean;
  };

  // Trust Score (0-100, algorithm-computed)
  trustScore: number;
  reviewCount: number;
  responseRate: number;     // % of inquiries answered
  completionRate: number;   // % of bookings completed
}

The AI agent discovery pipeline runs in four stages:

1. Intent parsing — extract service category, location, timing, budget from user request.
2. Candidate retrieval — all providers matching the category within radius, filtered by real-time availability.
3. Trust ranking — composite score: trust (40%), proximity (25%), availability (20%), price (15%).
4. Selection — top-ranked provider booked via platform API, or top three presented if user asked for options.

The provider subscription model has three tiers, each unlocking more of the machine-readable surface:

Tier	Core Value	Price	Target
Starter	Basic listing + AI discoverability	$29/mo	Solo operators
Growth	Priority ranking + analytics + bookings	$99/mo	Small businesses
Premium	Top placement + API access + multi-zone	$299/mo	Multi-location

The subscription revenue alone scales steeply. At 5,000 providers with a typical mix of 60% Starter, 30% Growth, 10% Premium, monthly recurring is around $90K. Per-booking fees ($1–$3 depending on tier) at typical volume add another $35K. Sponsored listings (premium placement in agent results, opt-in only) add the remainder.

The trust signal is the most defensible part of the architecture. Reviews on Google Maps are gameable — fake review farms produce five-star ratings at $5 each. AI agents need cryptographically verified reviews tied to completed transactions. The platform’s verified-review pattern signs each review with the customer’s DID at the moment of transaction completion; agents can verify the signature and assign weight accordingly. Manipulated reviews lose all weight; verified reviews carry the platform’s trust.

The recommendation engine uses four weighted signals:

• Trust signal (40%) — composite of verified reviews, completion rate, response rate, complaint resolution, credential status.
• Proximity signal (25%) — distance and travel time, weighted by service radius and zone.
• Availability signal (20%) — real-time slot availability matching user’s timing constraint.
• Context signal (15%) — service category fit, specialization match, price range alignment.

The weighting is configurable per platform instance. A medical referral platform might weight trust at 60% and proximity at 15%. A food delivery platform might weight availability at 40% and proximity at 35%. The book covers the tuning patterns for each vertical.

Where this scales

The article covers the AI-first discovery model and the recommendation engine. The book has the operational layers:

• MCP and A2A integration — how AI agents discover and consult the platform via Model Context Protocol tools, the agent-card declarations, and the A2A broadcast pattern for provider agents announcing real-time availability across the wider ecosystem.
• Trust and reviews in local commerce — verified reviews, trust badges from W3C verifiable credentials, dispute resolution, and the trust flywheel that makes manipulation economically unviable.
• Revenue projections — the $141K/month conservative model across 5,000 providers, with subscription tier mix assumptions, booking volume per provider, sponsored listing adoption rates, cost structure, and Year 2 expansion math.
• Launch playbook — city-by-city expansion, the 100 Providers program, density-before-breadth tactics, network effects at scale, and the 12-month execution timeline with month-by-month milestones.
• The data moat — how booking data improves recommendations recursively, voice-first local commerce patterns, and the path from “discovery platform” to “commerce operating system” for local services.

The book is built around the actual architecture of near.now, the local discovery layer in the Pragma.Vision ecosystem.

Included with the book

• local-business-ai-readiness-audit.md — a 25-point checklist for evaluating any local business against the five machine-readability requirements. Score the business, prioritize the gaps, and ship to AI-agent visibility in under a week.
• The $141K/month revenue model — full spreadsheet with subscription tier mix, booking volume, sponsored listings, cost structure. Adjust the per-city growth curve and see the consolidated path-to-revenue shift.
• MCP tool reference implementation — the near_now_find_provider and near_now_book_provider tools spec’d in JSON Schema, drop-in to a Cloudflare Workers MCP server.

Get the full picture

Readers of this also chose

Questions readers ask

How do you compete with Google when they have all the listings?

You don't compete on listing volume. You compete on machine-readability and real-time accuracy. Google Maps has millions of stale entries; the AI agent reads them as risk signals and discards them. A platform with 5,000 fully-verified, real-time-synced, cryptographically-signed providers outperforms Google for AI-agent traffic in that vertical and geography — even with two orders of magnitude fewer entries.

Why would a business pay $29/month when Google Business Profile is free?

Different value proposition. Google Business Profile gets the business found by humans on Google Maps. near.now's Starter tier gets the business found by AI agents (Claude, ChatGPT, Gemini, Siri) issuing local recommendations. As voice-activated search crosses 65% of local queries, AI agent visibility becomes the more valuable shelf. The book has the talking points for provider sales.

How long does the first city take to reach density?

The 100 Providers program is the supply-side milestone — typically 3–4 months from a cold start. Density is reached when the platform can serve 90% of category-specific demand from within-radius providers; this lands around 4–6 months in a typical mid-sized metro. The book has the city selection criteria (population, AI-adoption index, competitor presence) that shortens this.

What's the integration cost for a local business?

Self-serve onboarding takes a provider 20–30 minutes for the Starter tier. The platform pulls existing data from Google Business Profile (with the provider's permission) to pre-fill 70% of the structured fields. The remaining 30% (real-time availability, service radius zones, specialization tags) is what the platform adds. Growth and Premium tiers add deeper integrations (calendar sync, payment integration, multi-zone management) that take longer but unlock the higher-revenue features.

What's the refund policy?

Lemon Squeezy's standard refund window applies. If the playbook doesn't fit your local market, the refund link is in the receipt email.