Realty ONE Group Costa Rica · AI-native real-estate platform
Platform Feature Inventory & Investor Route-Map
A complete, honestly-graded map of everything the platform does — what is live in production, what is internal, and what is frontier R&D — with an approximate completion indicator for each capability and a phased route to close the gaps.
Neo Real Estate is a full-stack, AI-native real-estate operating system built for Costa Rica (a market with no unified MLS) and its US/Canada investor buyers, spanning 12 product pillars and 150+ discrete capabilities. The revenue-generating core is already live in production: a multi-tenant, three-language, four-currency consumer storefront with conversational AI search; an AI lead concierge ("Mia") with omnichannel capture, routing and follow-up; a full back-office CRM and real-time collaborative workspace used by an office of ~50+ agents; a $0-marginal-cost AI video/creative engine; an autonomous organic-marketing publisher; a complete sales-closings, commission-split and franchise-fee accounting engine; and a proprietary market-intelligence dataset (16 portals ingested, ~56K deduplicated for-sale listings; ~59K total index docs incl. rentals) rivals cannot quickly replicate. Underneath sits a mature, self-hosted infrastructure layer — local-first Electric SQL sync, Better Auth multi-tenant identity, an in-house Cloudflare-based asset/CDN platform (aquienpz), and postmortem-hardened deploy/observability tooling — that keeps marginal cost near zero as the platform scales. The frontier bets — realtime voice agents (proven in restaurants, not yet in real estate), iPhone-based 3D Gaussian-splat property tours, autonomous paid-ads spend, and a white-label "sell-the-platform" product (EstateKit) — are architected and partially built but honestly still experimental or throttle-held pending owner go/no-go decisions rather than engineering. Overall portfolio maturity is ~75% average completion, with the moat concentrated in owned data, owned low-cost infrastructure, and AI-native workflows that compound with usage. The nearest-term value unlocks are operational flips (connect the live WhatsApp channel, enable governed paid spend, ship the EstateKit real-data adapter), not greenfield builds.
159
Distinct capabilities
inventoried
inventoried
104
Live in production
24 consumer-facing
24 consumer-facing
133
Shipped or in active use
live + beta + in-progress
live + beta + in-progress
23
Frontier / planned
experimental + roadmap
experimental + roadmap
76%
Portfolio-wide
average completion
average completion
Status
Audience
Pillar 01
Consumer Storefront & Discovery
86%
maturity · 15 capabilities
The public-facing side is largely live in production: a multi-tenant property website across branded agent/office subdomains, in three languages and four display currencies, with app-like instant property views, map + conversational AI search, and shareable listings/reels. Buyer-account loops (favorites, compare, alerts, accounts) are live but still maturing; SEO/AEO discoverability and brand-neutral co-broke sharing are shipped.
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| Multi-tenant public property storefrontThe public property websites where buyers browse listings, agents, offices and developments — served on the main site and per-tenant/agent subdomains. | Public | Live | 95% | One codebase powers unlimited branded agent/office sites — a low-marginal-cost multi-tenant moat vs. one-site-per-agent rivals. |
| Instant property detail modal + deep linksTapping a property opens its full detail instantly as an overlay with no page reload, and every listing has a shareable direct link that opens straight into that view. | Public | Live | 95% | App-like instant navigation and clean deep links drive engagement and make every listing trivially shareable. |
| Multi-currency display (USD/CRC/EUR/CAD)Prices display in the visitor's currency (USD default, CRC, plus display-only EUR/CAD) via switcher and geo default, with a transaction-in-USD disclaimer. | Public | Live | 95% | Native multi-currency directly serves the US/Canada/Europe investor buyers driving Costa Rica's market growth. |
| Multi-language storefront (ES/EN/FR)The entire site, listing content and search served in neutral Spanish, English and Canadian French with proper SEO hreflang. | Public | Live | 95% | French captures Canada — CR's #2 source market — and the AI-maintained pipeline makes each new language cheap to add. |
| Map + list explore searchUnified map and list search of the full inventory — pan/zoom a clustered map, filter, and drive results from the map. | Public | Live | 90% | Map-first search is table stakes; owning the clustering + filtering stack keeps discovery fast and fully in-house. |
| Favorites & saved collectionsSigned-in buyers heart properties and organize saved listings into collections that persist across devices. | Public | Live | 90% | Saved listings create return visits and a first-party buyer profile — raw signal for retargeting, alerts and lead scoring. |
| Brandless broker co-broke market-shareAgents share a third-party market-search property as an unlisted, fully brand-neutral page on a separate neutral domain with proxied photos, honoring TTL/takedown. | B2B | Live | 90% | Enables the Costa Rica co-broke workflow legally and cleanly, expanding shareable inventory beyond own listings. |
| Compare properties + shareable comparisonAdd properties to a compare tray for a side-by-side spec comparison with AI assist, then share a frozen read-only snapshot with a client. | Public | Live | 88% | Compare-and-share turns the platform into the agent's decision-support tool, deepening both buyer and agent stickiness. |
| Conversational AI property search (Ask)Buyer-facing chat search — ask in plain Spanish/English and get streamed answers with real property cards inline; conversations can be shared and forked. | Public | Live | 85% | AI-native discovery is the platform's core differentiation and is answer-engine (AEO) ready; proactive-alerts extension still planned. |
| Shareable property pages & video reelsEvery listing has a rich public share page plus auto-generated short vertical video reels for social sharing. | Public | Beta | 80% | Turns every listing and agent into a zero-CAC distribution channel; the share-reel backend is still finishing. |
| SEO / GEO / AEO discoverability layerSitemaps, hreflang, JSON-LD structured data, robots and an llms.txt so both search engines and AI answer engines surface the listings. | Infra | Live | 80% | Answer-engine optimization positions the inventory to be cited by AI search — an emerging, under-contested acquisition channel. |
| Price alerts & saved searchesBuyers save a search or property and get notified on price changes and new matches, via a dedicated alerts page and notify-me prompts. | Public | Beta | 78% | The highest-intent re-engagement loop in real estate; smarter-matching upgrade is the next conversion lever. |
| Editorial blog, area guides & comparison guides (public SEO)Public long-form content — a blog, neighborhood/area narratives and side-by-side comparison guides — that brings organic search and AI-answer-engine traffic to the listings. | Public | Live | 78% | A compounding, owned acquisition channel: content the AI helps produce ranks and feeds the same buyer funnel at zero media cost. |
| AI article-to-audio narrationTurns written guides/articles into natural-voice audio (text-to-speech) so content is consumable as a podcast-style listen. | Public | Live | 76% | Repurposes one piece of content across text and audio channels automatically — more reach per unit of content, no studio. |
| Consumer accounts & social loginEnd buyers create an account and sign in (incl. social login) to unlock favorites, collections and alerts, with sessions working across all tenant subdomains. | Public | Beta | 70% | A consumer identity layer is the foundation for owning the buyer relationship rather than renting it from portals. |
Pillar 02
ONE CRM & Leads Engine (Mia Concierge)
87%
maturity · 18 capabilities
The revenue conversion engine: turns traffic into qualified, followed-up, attributed leads. Lead capture, the AI qualifier (Mia), the leads pipeline, a unified multi-channel inbox, routing/handoff, follow-up cadences, consent enforcement, the agent daily loop, and intelligence dashboards are built and live (Instagram is a live channel today). The single biggest open gap is connecting the live WhatsApp Business channel — dominant in Costa Rica — where code is complete but the WABA connection and template sign-off await an owner go-live decision.
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| Legacy storefront realty-chat widget (superseded)The original anonymous visitor-to-agent storefront chat, intentionally retired and replaced by Mia + the new /crm surface. | Public | Retired | 100% | Listed for honesty: validated the realtime chat + takeover patterns Mia now productionizes; no longer a live consumer surface. |
| Chatwoot support inbox (third-party, being displaced)A production Chatwoot instance still handles some support conversations, progressively being folded into the in-house omnichannel inbox. | Internal | Live | 100% | Listed for transparency: a licensed third-party tool in use today, on a path to replacement by the owned inbox (removing a recurring SaaS cost). |
| Storefront lead capture + showcase attributionEvery property/share-link contact form is captured instantly and routed to the right advisor, correctly crediting the agent who shared the link. | Public | Live | 95% | Correct commission attribution on shared links protects agent trust — the foundation the success-fee lead economics rests on. |
| Focalized property-inquiry funnel (fail-safe)A phone-first inquiry form with intent chips that captures the lead even if the user abandons midway, so partial inquiries are never lost. | Public | Live | 95% | CRO-tuned capture with abandon-recovery lifts top-of-funnel yield from the same ad/SEO traffic. |
| Mia — AI lead-qualifier conciergeAn always-on AI assistant that greets, answers and qualifies each lead in the visitor's language, scoring intent before a human steps in. | Public | Live | 92% | Always-on AI concierge in a ~91% WhatsApp-penetration market is the structural moat — sub-5-minute contact is ~21x more likely to qualify. |
| Speed-to-lead watchdogs + auto-acknowledgementBackground timers fire an SLA alert if a new lead isn't worked fast, auto-enroll stalled leads, send instant auto-acknowledgement, and log every tap as a touch. | Internal | Live | 92% | Turns the 'contact within 5 minutes' research edge into enforced system behavior rather than agent discipline. |
| Agent daily follow-up loop (queue, snooze, Hot Sheet)A phone-first daily workflow: a server-sorted 'what to do now' queue with one-tap snooze, a morning Hot Sheet digest and push notifications. | Internal | Live | 92% | The computed do-or-snooze queue is what makes the automation actually get worked — it operationalizes agent productivity. |
| ONE CRM leads pipeline (kanban)A visual drag-and-drop pipeline of leads by stage with per-agent scoping, scoring, hot/warm/cold temperature, archive, merge and reassignment — deliberately simplified for non-technical agents. | Internal | Live | 90% | Purpose-built CRM benchmarked vs Attio/Follow Up Boss/Lofty; the 'calm UI' investment targets the #1 reason real-estate CRMs fail — agent adoption. |
| Lead routing & agent handoff (claim + SLA reroute)Leads route to the right advisor; humans take over from Mia anytime; countdown claim-offers let idle leads be grabbed; unanswered leads auto-reroute on SLA expiry. | Internal | Live | 90% | Automated reassignment ensures no lead sits unanswered — protecting the answer-rate SLAs pay-at-closing referral programs enforce. |
| Consent & compliance floor (STOP / do-not-contact)A lead texting STOP is enforced everywhere instantly: a revocation is written and a hard do-not-contact flag blocks every outbound send surface. | Infra | Live | 90% | A working opt-out/consent boundary is the legal prerequisite for turning on paid automated outreach at scale without regulatory risk. |
| Contact identity resolution, dedup & enrichmentAutomatically merges duplicate contacts across channels into one record (email/phone/handle) with a human review queue, plus scoring and enrichment. | Infra | Live | 90% | Clean de-duplicated identity is the substrate every other feature (routing, scoring, attribution, AI context) depends on. |
| Provider-agnostic omnichannel engine + IG auto-replyMia talks across channels through a pluggable adapter design (HighLevel first); she currently auto-replies to Instagram DMs in production. | Public | Live | 85% | Provider-agnostic design avoids vendor lock-in and lets the same AI brain add channels as an adapter, not a rebuild. |
| Unified omnichannel CRM inboxOne back-office inbox showing every lead conversation with merged identity, qualification tags, channel badges, per-message translation, AI-takeover and a live 'who's online' roster. | Internal | Live | 85% | Keeps agents inside the product instead of leaking to personal phones; live for Instagram today, WhatsApp/SMS/email code-complete but not yet connected. |
| Lead intelligence dashboards & reportingBroker analytics across 8 report types — stalled-lead, speed-to-lead-by-source, source-ROI, agent leaderboard — with an AI-written narrative summary. | Internal | Live | 85% | Source-ROI + closing-attribution reporting proves the success-fee model and optimizes ad spend toward closings; live alerting still to come. |
| Follow-up cadence / nurture engineAutomated multi-step follow-up sequences (incl. sub-day hot-burst cadences) that pause the moment a lead replies and respect frequency caps and quiet hours. | Internal | Beta | 80% | Auto-pause-on-reply is the #1 gap vs FUB Action Plans; machinery is built and tested — awaiting an owner consent-gate go-live flip, not engineering. |
| Leads/CRM MCP server + Mia broker toolsetA secure, owner-scoped connector letting the broker's own AI read the pipeline (360 view, stalled leads, source-ROI) plus Mia's internal follow-up tools. | Internal | Beta | 78% | Exposing the CRM as callable AI tools makes the broker's own AI a pipeline supervisor — a defensible surface as ops go AI-native; mutations still v2. |
| Newsletter capture & email-marketing audiencesCaptures newsletter subscribers into managed email audiences (Resend) for ongoing nurture and announcements. | Public | Beta | 72% | A first-party email list is a durable, platform-independent marketing asset — owned reach that survives any ad-platform change. |
| WhatsApp channel + template registryIn-product WhatsApp with a reusable template library and a 24h-window-aware composer, so WhatsApp threads become CRM records rather than a link-out. | Public | In progress | 55% | WhatsApp is the dominant CR channel and the funnel's economic core; code is ready but the live WABA connection is the single biggest remaining gap. |
Pillar 03
Sales Closings, Commissions & Franchise Accounting
83%
maturity · 8 capabilities
The platform's money engine — a complete, shipped system (doc 36, 12 workstreams merged) that records a sale from contract to disbursement and does the commission math on top. It covers the closing/CIF wizard, agent/office/franchise commission-split calculation, the deposit→payout payment lifecycle, a broker/owner accounting operations dashboard, Top-Producers leaderboards, and — critically for a franchise — automated office→franchise monthly fee reporting and FranConnect statistics. This is the most directly revenue- and unit-economics-relevant system in the whole platform and it is largely live.
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| Closing / CIF wizard (record a sale)A guided flow to register a closed sale — property, parties, price, dates and the CIF paperwork — turning a handshake into a structured, auditable transaction record. | Internal | Live | 88% | The system of record for revenue events — every commission, report and leaderboard downstream is computed from these closings. |
| Commission-split & payout math (agent / office / franchise)Automatically computes how each closing's commission is divided between the agent, the office and the franchise, including referral and external-agent splits. | Infra | Live | 88% | Getting split math provably right is the hard, trust-critical core of a brokerage's finances — and it is automated, not spreadsheet-driven. |
| Top-Producers leaderboardsRanks agents by closed volume and commission over a period, powering recognition and healthy competition. | Internal | Live | 88% | Ties the shipped analytics to real closed revenue (not vanity activity) — a retention and motivation lever grounded in money. |
| Payment lifecycle tracking (deposit → disbursement)Tracks the money through its stages — earnest deposit, funds received, commission disbursed — so nothing falls through the cracks between contract and payout. | Internal | Live | 82% | Cash-flow visibility on in-flight deals is what turns a CRM into a financial operating system for the brokerage. |
| Broker / owner accounting operations dashboardA management console showing closings, revenue, outstanding payments and commission liabilities across the office in one place. | Internal | Live | 82% | The owner-facing P&L view — the surface an investor or franchise principal actually uses to run the business. |
| Office → franchise monthly fee reporting + FranConnect statisticsAutomatically produces the monthly numbers a franchise office owes and must report upward, in the format the franchisor's FranConnect system expects. | B2B | Live | 80% | Automating franchise compliance reporting is a concrete wedge to sell the platform to other Realty ONE Group offices — it removes their monthly accounting pain. |
| Per-listing charity contributionOptionally allocates a portion of a closing's proceeds to a chosen charity, tracked through the accounting system. | Internal | Live | 78% | A brand/ESG differentiator wired directly into the money flow rather than bolted on — resonant with the Realty ONE Group community ethos. |
| External-agent (referral) portal & payoutsA scoped surface for outside/referral agents to see their attributed deals and expected payouts without accessing the full back-office. | Internal | Beta | 74% | Extends the commission engine to the referral network — the plumbing for pay-at-closing partnerships that expand deal flow. |
Pillar 04
Back-Office, CRM & Team Collaboration
75%
maturity · 14 capabilities
The internal operating system ~50+ agents and admins use daily in production. Real-time collaborative listing creation, listing management + audit history, internal team chat, agent profiles and performance dashboards, and admin oversight tools (catalog health, shared media folders, org-wide draft oversight) are shipped and mature. A smaller set (chat 'stories', monthly ranking emails) is planned; the collaborative blog editor is in beta.
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| Collaborative Draft Listing Wizard (real-time co-editing)A step-by-step wizard where an agent, an AI assistant, or a second teammate build a listing together in real time — edits, uploads and AI suggestions sync instantly and survive refresh/offline. | Internal | Live | 95% | Google-Docs-grade collaboration on the core revenue object; the CRDT foundation is reused across chat, blog and kanban — a durable technical moat. |
| Listing History / Bitácora (audit timeline)A GitHub-style timeline on every property showing who archived, re-published, re-priced or edited it and when. | Internal | Live | 92% | Trust and compliance layer — an immutable per-asset event log, foundational for future auditability and franchise governance. |
| Listing Management workspaceThe central place agents manage properties: virtualized grid/table views, active/inactive, sort/filter, bulk actions and quick edits, tuned for the installed phone app. | Internal | Live | 90% | Mature, mobile-first daily driver for the sales team, performance-hardened for the PWA where most usage happens. |
| Agent Profiles with per-field privacy controlsEditable agent profiles where each agent can hide individual fields from public pages via a simple toggle, enforced server-side on every public surface. | Internal | Live | 90% | Agent-respecting privacy that propagates to consumer sites — reduces friction in agent onboarding and retention. |
| Catalog Health admin dashboardAn admin control room showing catalog + crawler health — freshness per source, dead-photo rate, geo/price coverage, dedup stats. | Internal | Live | 90% | Operational observability for the data asset that powers the product; catches ingestion regressions before they reach customers. |
| Admin Media Folders (shared media library)Google-Photos-style shared folders of original-quality photos/videos, link-shared: anyone can view/upload, only the admin can download/rename/delete. | Internal | Live | 90% | Removes a real operational pain (sharing raw shoot files) inside the platform instead of Google Drive/WeTransfer — assets stay in-house. |
| Agent Performance dashboard + shareable weekly cardEach agent sees real numbers (views, contacts, shares, top listing) with a team leaderboard rank and can generate on-brand shareable art to post on social. | Internal | Live | 88% | Honest engaged-only metrics build agent trust; the shareable card turns agents into a zero-CAC distribution channel. |
| Omniscient AI assistant memory & activity systemThe back-office AI's long-term brain: it remembers each user's preferences in plain language and logs activity, injecting relevant context into every conversation. | Infra | Live | 88% | The personalization/memory layer for an AI-native workspace; compounds with usage and is reused across every AI surface. |
| Admin org-wide Draft oversightAn admin console to see, open and edit every draft across the office so managers can unblock or finish any agent's work-in-progress. | Internal | Live | 85% | Manager-tier control surface; scales the collaboration model from individual to office oversight. |
| Internal Team Chat (DMs + channels)In-app 1:1 DMs and admin-created channels with live presence, typing indicators, reactions and reliable catch-up after reconnect, all inside the PWA. | Internal | Live | 85% | Keeps team communication and its data inside the platform rather than WhatsApp, reusing the proven realtime stack at near-zero marginal cost. |
| Collaborative AI Editorial / Blog editorA real-time, multi-author blog/guides editor where admin and agent co-write from different devices while a Grammarly-style AI peer drafts and suggests edits live. | Internal | Beta | 75% | Reuses the CRDT engine for SEO content production, turning a growth channel into a team+AI activity rather than a bottleneck. |
| Quotes / estimates surfaceA back-office surface to build and track client-facing quotes/estimates tied to listings. | Internal | Beta | 65% | Minor but real operational surface; keeps quoting inside the platform's system of record. |
| Team Chat Stories / Estados (ephemeral 24h)WhatsApp-style ephemeral 24-hour status posts for the team with seen-by and reactions, plus composer fluidity polish. | Internal | Planned | 12% | Roadmap engagement feature; low priority relative to the shipped chat core. |
| Monthly performance reports & Top Producers rankingAutomated month-start emails: a personal performance card to each active agent plus a shareable 'Top Producers del mes' ranking for management. | Internal | Planned | 10% | Extends shipped weekly analytics into recurring outbound engagement; near-term, low build risk (reuses agent-performance contract). |
Pillar 05
AI Creative Studio & Video Engine
86%
maturity · 16 capabilities
Turns a listing's photos into a polished, branded social video with almost no human effort — an AI 'Creative Director' picks photos, writes the script/captions, choreographs motion and scores music, rendered on a $0 in-house engine. Most of the system is live in the back office; the newest polish (text-behind-subject captions, motion captions, music intelligence, real-footage tours) is shipping/rolling out. A premium generative-video path (Veo/Kling/Seedance/Wan) is available but parked in favor of the zero-cost engine.
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| Hyperframes render engine — RETIREDThe original browser-based (Chromium/GSAP) rendering engine, now archived; all production rendering runs on the faster/cheaper Takumi engine. | Infra | Retired | 100% | Listed for honesty: dead code preserved for rollback — the platform consolidated onto one cheaper engine. |
| Deterministic $0 render engine (Takumi + FFmpeg)The in-house engine that assembles photo slideshows into finished MP4s in seconds with no per-second AI billing — the everyday workhorse. | Infra | Live | 95% | Cost moat: renders at ~$0 marginal cost vs per-second generative billing — a structural margin advantage enabling unlimited content. |
| Adaptive HLS reel playback on public storefrontOn public listing pages, reels stream adaptively (240p–1080p) with an instant real-frame poster instead of a heavy single file. | Public | Live | 92% | Consumer-facing performance win: fast mobile playback lifts engagement and time-on-page for buyer traffic. |
| AI Video Studio (listing → social video)End-to-end: an agent picks a listing and the system produces a ready-to-post branded vertical video from its photos, with captions, music and per-platform sizing. | Internal | Live | 88% | The flagship: collapses a costly agency/videographer workflow into a self-serve tap, generating scaled social content. |
| Photo enhancement & upscalingCleans up and upscales listing photos before they enter a video so low-quality sources still produce sharp output. | Internal | Live | 88% | Raises the floor on input quality across a heterogeneous library — makes the content engine usable on real, imperfect listings. |
| AI Creative Director (auto shot-planning, photo selection & scripting)An AI 'director' chooses the best photos, orders them into a narrative, writes the on-screen script/hooks, and plans pacing so a non-expert gets a good video. | Internal | Live | 85% | The intelligence moat — quality/taste is automated, so output is consistently on-brand and conversion-oriented at scale. |
| ONE's Video Creator wizard (CapCut-simple UI)A guided, jargon-free wizard that lets a non-technical agent make a video photos-first, with AI-suggested styles and visual previews. | Internal | Live | 85% | Adoption lever: usability determines whether every agent creates content, not just the tech-savvy few — key to content volume. |
| Property 'Share Video' reel (branded 9:16)From a listing page, an agent generates and shares a short branded vertical slideshow sized for Stories/Reels/TikTok/WhatsApp. | Internal | Live | 85% | Turns every listing into instantly shareable social content — a viral-distribution primitive at the point of daily agent work. |
| Motion-path captions (kinetic text)Named animated caption motions (slide, rise, motion-path) tuned for the high-volume TikTok/Reels register. | Internal | Live | 85% | Kinetic captions measurably lift short-form retention, on the zero-cost render path. |
| Real-footage video creator (Media Enhance)Builds videos from actual property video footage (not just photos), producing multi-language 16:9 reels with a brand-only outro. | Internal | Live | 85% | Extends the engine beyond stills to walk-through footage and three languages — broadening content types and market reach. |
| Per-platform delivery & format variantsAutomatically outputs the right size/quality per network — 1080p master for TikTok/IG/FB and a lighter 720p tuned to survive WhatsApp compression. | Internal | Live | 85% | Removes per-platform re-editing friction and prevents quality loss on WhatsApp — a primary LATAM channel. |
| Personal-brand caption stylesA picker of caption/typography styles so an agent's videos carry a consistent personal or team brand across all content. | Internal | Live | 82% | Personalization at scale — agents build their own brand while staying inside franchise guardrails. |
| Music: curated sound library + AI music directorAutomatically scores each video with a fitting soundtrack from a curated, licensed library, syncing the track to the video's beats. | Internal | Live | 80% | Solves the #1 licensing/copyright risk in social video with a compliant, auto-selected soundtrack (Lyria fresh-bed gen gated off). |
| Reel quality gate, video QC & review cardsAutomated checks reject blurry/pixelated photos and flag bad renders, plus a review workflow where the AI director learns from rejections. | Internal | Live | 80% | Closed-loop quality control — the system improves over time and protects brand consistency without manual QA headcount. |
| Multi-provider generative AI video (Veo/Kling/Seedance/Wan)Premium path that turns a still listing photo into a moving cinematic clip using top image-to-video models, with automatic provider fallback. | Infra | Beta | 78% | Optionality + differentiation: keeps the highest-end cinematic capability provider-agnostic; currently parked in favor of the $0 engine. |
| Text-behind-object (subject-occluded) captionsThe CapCut/VEED-style effect where a caption appears behind a person or foreground object so the subject stands in front of the text. | Internal | Beta | 78% | Visual parity with the best consumer social-video editors — output looks native/high-end, not templated. |
Pillar 06
Autonomous Marketing & Growth Engine
77%
maturity · 14 capabilities
A code-governed system that markets the company at near-zero marginal cost: an autonomous 'Marketing Director' decides what/when/where to publish, a multi-format publisher is live across 3 Facebook Pages and 3 Instagram accounts (posting daily, currently under human approval before flipping fully autonomous), a Telegram bot hands agents a fresh reel daily, and a governed paid-ads rail is fully built and staged but deliberately spending $0 pending an owner budget decision. TikTok/YouTube publishers are code-complete but audit-gated. Honest headline: engine assembled and running organically, throttle held on paid spend.
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| Full-funnel attribution (utm/gclid/referrer)Every lead — form, chat, share page, valuation funnel — is stamped with its true marketing source, campaign, country and a test-vs-real flag. | Infra | Live | 90% | 'No spend before attribution' — every future marketing dollar is measurable and the company can build its own CR cost-per-lead benchmarks. |
| Multi-zone brand routing (CR / Surf / Executive)Automatically sends each listing to the right brand account — general CR, beach/Surf, or luxury Executive — across matching FB Page and IG profile. | Infra | Live | 88% | One engine runs three differentiated brand voices from a single codebase — the structure to scale zones or white-label other offices. |
| Native multi-format render matrix (per-placement quality)Every post is composed natively for its slot (9:16, 4:5, 1:1, 16:9) with platform-safe margins and designed covers — not one video cropped everywhere. | Infra | Live | 88% | Native quality at $0 marginal cost is the structural edge — creative volume is free while rivals pay per asset. |
| Multi-platform native publisher (IG + FB, multi-format)Publishes to Instagram and Facebook in every native format — reels, 4:5 feed video, photos, carousels, stories, link posts — automatically after approval. | Internal | Live | 85% | Official platform APIs only (no fragile proxies), durable against account bans — owned, compliant plumbing rather than a growth hack. |
| Ad-creative factory + licensed audio engineTurns listings into ad-ready short videos with tracked click-through links, hook presets, per-creative campaign tags and professionally licensed music cleared for paid ads. | Infra | Live | 85% | The same UTM-tagged, licensed asset feeds both organic and paid funnels — measurement and monetization are built in from creation. |
| Seller-acquisition valuation funnel (¿Cuánto vale su propiedad?)A consumer landing page where a homeowner gets an instant, consumer-safe value estimate and becomes a seller lead. | Public | Live | 85% | No CR competitor runs an instant-valuation seller funnel — one captured listing is ~$5-15k in commission, potentially paying for the whole marketing plan. |
| Autonomous Marketing Director (organic publishing brain)An automated director that decides which listings/reels to publish, in what order and at what optimal time, fairly rotating across agents on a schedule. | Internal | Live | 82% | Code owns fairness/cadence/timing; the LLM is a bounded, traced re-ranker that can never break an invariant — 'safe autonomy'. Currently owner-supervised. |
| Telegram agent reel distribution botEach agent gets a fresh personalized reel in Telegram daily and taps like/dislike; liked reels become shareable and a group channel pushes tutorials. | Internal | Live | 80% | Distributes content through the agent network with zero manual work, and like/dislike feedback trains the Director. Early pilot, rolling out. |
| Marketing Studio cockpit (J.A.R.V.I.S.)A back-office command center where staff see what the brain is planning/publishing, inspect each post's 'ingredients' (music, clips, licensing, cost), and issue commands. | Internal | Live | 80% | The human-oversight and explainability layer that makes autonomy trustworthy and demoable — an investor can watch the brain work. |
| Studio MCP + weekly automated batch routineAn automation server plus a scheduled AI routine that each week proposes and renders a batch of reels from top listings and notifies the owner with preview links. | Infra | Live | 80% | Removes the last manual step from the organic engine — content proposes and renders itself weekly, with owner review the only gate. |
| Cross-network growth loop + follower KPI dashboardCross-promotes Instagram content to Facebook and captures daily follower-count snapshots per account, surfaced in a back-office growth dashboard vs a baseline. | Internal | Beta | 70% | Uses larger IG followings to grow smaller FB pages and gives investors a real, measured follower-growth chart per brand. |
| Governed Meta paid-ads growth railA budget-safe system to run FB/IG follower-growth and awareness campaigns per brand zone, with hard spending caps, always-paused-on-create and approval keys. | Internal | In progress | 68% | Fully built and verified live against Meta's API but deliberately spending $0 — the throttle is an owner budget decision, not missing code. |
| Boost-winners loop (organic-to-paid amplification)Detects which organic posts win by real engagement and proposes a small paid budget behind them, with owner approval before any spend. | Internal | In progress | 55% | Only pay for content that already proved organic traction — capital-efficient by design; built and tested but running dry until the paid rail is enabled. |
| TikTok + YouTube Shorts publishers (audit-gated)Publishing modules for TikTok (Spark Ads + direct) and YouTube Shorts, built and wired but restricted to private/self-only until platform app-review audits clear. | Internal | Experimental | 42% | Channel expansion is code-complete and waiting on external platform audits, not engineering — a low-cost future unlock. |
Pillar 07
Market Intelligence & Data Moat
78%
maturity · 15 capabilities
The company's data moat: an automated pipeline continuously crawls 16 Costa Rican sources, deduplicates them into a single canonical inventory (~56K deduplicated for-sale listings; ~59K total index docs incl. rentals), and turns that into pricing intelligence, a heatmap-driven market search, listing-quality scoring, market analytics and public market reports. Most is live and used internally today; forward-looking pieces (long-horizon price indices, lifecycle analytics, AI-callable inventory, developments pricing) are data-accumulating or on the roadmap. Defensibility comes from owning a clean, deduplicated, source-attributed view of a fragmented, MLS-less market. (Honest caveats: several of the 16 portals share one WASI/MLS backend, so independent feeds number ~12; and the current corpus scale accrued rapidly over ~2 weeks — it is a fresh, self-refreshing asset, not a long-accumulated one.)
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| Buscador — back-office market search (map + clustering + facets)An internal tool showing the whole market on a clustered map with source and rich filters — agents search the entire market, not just company listings. | Internal | Live | 92% | Gives every agent a market-wide search advantage over consumers and rival brokerages that only see their own book. |
| Price Heatmap — 'Zonas de Calor'An Uber-fare-style heatmap of price-per-square-meter across zones so users instantly see where a market is expensive or cheap. | Internal | Live | 90% | Visualized price intelligence is an instantly demonstrable, sticky feature and the visible tip of the data moat. |
| Listing Quality ScoreAutomatically grades each listing's completeness and quality (photos, description, watermarks, missing fields) so agents know what to fix. | Internal | Live | 90% | Systematizes listing quality at scale, improving conversion and creating a measurable operational metric for the brokerage. |
| Public Market Report (Reporte de Mercado)A branded, shareable public web report of prices by region, property type and bedrooms, plus yields — a consumer- and lead-facing market study. | Public | Live | 90% | A credible public market study built on the proprietary dataset — a lead-gen and brand-authority asset no competitor can match. |
| Multi-source market ingestion pipeline (16 sources)Automatically and continuously collects for-sale and rental listings from 16 Costa Rican real-estate sites/portals into one central database. | Infra | Live | 88% | In a market with no unified MLS, owning a broad always-fresh cross-portal feed is the core data moat rivals can't quickly replicate (note: 16 portals shown, ~12 independent backends — several share one WASI/MLS source). |
| Cross-source deduplication & canonical resolutionDetects when the same property is listed on multiple sites and merges them into one clean record so the market view isn't inflated by duplicates. | Infra | Live | 85% | Clean deduplicated inventory turns raw scraped data into a trustworthy pricing dataset — the hard, defensible part of a data business. |
| Search Analytics & BI dashboardAn internal dashboard tracking how people search — classic vs AI search vs the Mia concierge — with time-series and breakdowns. | Internal | Live | 85% | Proves which AI surfaces drive engagement; cross-surface attribution is a known remaining gap. |
| Market Data Health — freshness, sales-tracking & photo livenessBackground systems detect when listings disappear or sell, track price changes over time and prune dead photos — keeping the dataset honest and fresh. | Infra | Live | 85% | Data decays fast; automated freshness and self-healing keeps the moat from silently rotting — an under-appreciated defensibility layer. |
| Ask / external-market rich tool (AI concierge market search)Lets the AI concierge surface real listings from across the whole market (not just company inventory) directly in chat as rich cards. | Public | Live | 85% | Bridges the AI-native front door to the proprietary market dataset — a differentiated experience rival chatbots can't offer. |
| CMA / comp-anchored price recommendation engineProduces a defensible suggested price for a property by comparing it against similar recent listings (a comparative market analysis). | Internal | Live | 80% | Turns raw data into direct revenue leverage — better-priced listings sell faster — and is hard to copy without the underlying comp dataset. |
| New-development (obra nueva) listings — live product surfaceMulti-unit new-construction projects are a live listing type across the storefront and Mia — with promotion, AI enrichment and revalidation — distinct from the (planned) continuous pricing-sync pipeline below. | Public | Live | 80% | Captures the high-value new-construction segment developers pay to market; live today, with the continuous auto-sync pipeline as the roadmap upgrade. |
| Market Pulse — lifecycle analytics (DOM, absorption, MOI, temperature)Turns the live feed into decision metrics: days-on-market, absorption rate, months of inventory, sell-through and a market-temperature score. | Internal | Beta | 70% | Machinery is built and the clock is running; once ~60-90 days of history accrues it yields metrics normally only large MLS operators possess. |
| Listings MCP Server (AI-callable inventory + Apps SDK widget)Exposes the live property catalog as tools any AI assistant can call, plus a rich widget — positioning the inventory to appear inside ChatGPT-style apps. | B2B | Beta | 65% | A forward bet on AI-assistant distribution — makes the inventory reachable where buyers increasingly start their search. |
| Judicial auctions / remates surfaceA dedicated Costa-Rica-specific view of distressed and court-auctioned properties, resolved from public finca records into the canonical inventory. | Public | Beta | 65% | A differentiated distressed-asset product no consumer portal in CR offers cleanly — extends the data moat into a niche with motivated buyers. |
| Developments ingestion & pricing engineA pipeline to keep multi-unit new-development projects (pricing, units, availability) continuously in sync, not just created once. | Internal | Planned | 15% | Extends the data moat into the high-value new-construction segment; on the roadmap — upside, not a current asset. |
Pillar 08
Mia Voice Agent Platform
47%
maturity · 12 capabilities
A realtime, interruptible voice assistant on self-hosted LiveKit + Google Gemini 2.5 Live audio. What is actually live today is a restaurant voice-ordering concierge (one brand in production, one demo), each isolated as a web app talking to its own agent. The engine is deliberately multi-vertical/white-label — new brands plug in via a profile + catalog JSON — and early supermarket/hardware verticals exist. Native clients (desktop wake-word, mobile) and the real-estate application of Mia's voice remain prototypes/stubs. Honest headline: strong reusable voice platform with real production usage in restaurants, but the real-estate voice use-case is not yet shipped.
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| Realtime voice agent backend (LiveKit + Gemini 2.5 Live)The brain of Mia: a Python agent holding a natural, spoken two-way conversation in real time, letting the user interrupt mid-sentence (barge-in) like a person. | Infra | Live | 85% | Provider-agnostic realtime voice stack (self-hosted LiveKit avoids per-minute lock-in); barge-in + native-audio quality is the hard part and it works in prod. |
| Web push-to-talk client (restaurant)The public web page where a diner opens their phone browser, taps to talk to Mia, and orders out loud — no app install, no login. | Public | Live | 85% | Zero-friction consumer entry point that proves the voice UX to real customers today. |
| In-conversation dynamic product cards (OpenUI Lang)As Mia mentions a dish or item by name, its photo and details pop up on screen automatically so the customer sees what they're ordering. | Public | Live | 80% | Voice + synchronized visuals is a genuine differentiator over voice-only assistants; removes ordering ambiguity. |
| Voice ordering & cart tools (add/remove/confirm, kitchen panel)Mia builds the order by voice — add, remove, confirm — and confirmed orders move to a kitchen view. | Public | Beta | 68% | Moves Mia from chatbot to transaction engine; the remaining kitchen-lifecycle/POS work is the wedge into paid operations integrations. |
| Multi-vertical / white-label agent architectureOne codebase, many brands and industries: each client is a profile + a menu/catalog file, so a new business launches without touching anything live. | B2B | Beta | 65% | The real business asset — a repeatable, sellable voice-concierge platform for any catalog business, not a one-off restaurant app. |
| 'Ey Mia' wake-word model (hands-free activation)A trained voice-trigger so a user can say 'Ey Mia' to wake the assistant hands-free, tuned to Costa Rican speech. | Infra | Beta | 55% | Custom locally-accented wake word is defensible IP, but its value is unrealized until a native client ships it to users. |
| Supermarket voice vertical (aquienpz)An adaptation of Mia to help shoppers query a supermarket's assistant by voice, reusing the same engine as the restaurant version. | B2B | Experimental | 45% | First proof the platform generalizes beyond restaurants — validates the multi-vertical thesis with a second industry. |
| Hardware / device voice verticalAn early adaptation of Mia for a hardware/device context, resolving a different assistant profile per session. | B2B | Experimental | 30% | Signals ambition toward embedded/device voice but earliest stage — treat as R&D, not a product. |
| Native mobile app (Expo, iOS/Android)A planned phone app for talking to Mia natively (with wake word) instead of through a browser. | Public | Experimental | 20% | Not a shipping channel yet; the browser client covers reach today, so mobile is optionality rather than current capability. |
| Native desktop wake-word clientA planned always-listening desktop app that waits for 'Ey Mia' and then connects the voice session. | Public | Experimental | 15% | Earliest-stage client; the only surface that would use the trained wake word, so both are gated on the same future push. |
| Back-office real-estate voice assistant (webapp)An earlier in-app voice+text assistant that let agents navigate the back-office by talking — built and wired but removed because no one used it. | Internal | Retired | 10% | Do NOT present as live — it shipped, found no adoption, and was archived; proof the team can build in-app voice, not a current product. |
| Real-estate voice vertical of MiaThe intended application of Mia's voice to real estate (voice property search / agent assistant) — the reason this pillar matters for the company. | Internal | Planned | 8% | Honest gap: the voice platform is proven in restaurants but the core real-estate use-case is not built — a fast follow given the reusable architecture. |
Pillar 09
3D Capture & Immersive Property Tours (R&D)
47%
maturity · 11 capabilities
An R&D 'lab' letting an ordinary agent walk through a property with an iPhone and get back a photorealistic, walkable 3D tour (Matterport-style, no special hardware) using the phone's LiDAR/AR sensors plus cloud Gaussian-splat reconstruction. The full loop has been proven end-to-end on real rooms (live demo tours exist) and a capture app is on TestFlight, but it is early/experimental: not rolled out to users, no production URL, and quality has known limits (mirrors/glass). The strategic bet is that the foolproof guided-capture UX is the durable moat while viewer and splat engine are commodity open-source.
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| Guided iPhone 3D capture ('paint the room' HUD)A foolproof camera experience coaching an agent to scan a room by 'painting' every surface with a coverage ring, so a non-technical user captures a good scan first try. | Internal | Experimental | 58% | The capture UX is the explicit moat: making a good scan foolproof for non-expert LatAm agents is the hard, defensible part rivals underinvest in. |
| Dedicated capture-only iOS app (TestFlight)A clean standalone iPhone app focused purely on capture-to-tour, with guest mode, currently distributed via Apple TestFlight for internal testing. | Internal | Beta | 55% | An independently-releasable surface lets the capture experience iterate and reach real agents fast, decoupled from the main app's release cycle. |
| Cloud 3D reconstruction pipeline (COLMAP + Gaussian Splatting)The GPU pipeline that turns a phone scan into a 3D model: solves camera positions and trains a Gaussian-splat scene on a cloud A100, producing a photorealistic reconstruction. | Infra | Experimental | 55% | Built entirely on commercially-clean open-source on the company's own GPU quota — an ownable pipeline, not a per-scan fee to Matterport/Polycam. |
| Real-time capture coaching & quality oracleLive on-screen feedback during the scan ('move slower', 'avoid mirrors') driven by a scoreboard that tells a good capture from a bad one before any cloud processing. | Internal | Experimental | 52% | Turns reconstruction failures (~$2-3 of GPU per bad job) into pre-flight coaching — protects unit economics and feeds a capture-telemetry flywheel. |
| Splat compression (SOG) + walkable collision generationPost-processing that shrinks the model into a mobile-friendly format and builds an invisible floor/walls mesh so a viewer can walk through without falling through. | Infra | Experimental | 52% | The de-risked headless path means this runs automatically in the cloud job — removing the last Mac-dependent manual step. |
| Automated recon orchestration backend (upload → queue → GPU job)Server-side plumbing that receives a scan securely, queues it and launches the cloud GPU job automatically, with idempotency so a retry never double-pays for a run. | Infra | Experimental | 50% | Turns a run-a-script prototype into an automated, idempotent service — the step that makes 'tap a button and a tour appears' possible at scale. |
| LiDAR depth capture + reflective-surface maskingUses the iPhone's LiDAR to capture depth alongside video, improving reconstruction and auto-flagging mirrors/glass (the #1 thing that breaks 3D scans). | Infra | Experimental | 48% | Exploits hardware most competitors ignore to solve the hardest 3DGS failure case cheaply, without paying for server-side segmentation. |
| Walkable & dollhouse 3D tour viewerThe interactive browser tour: a first-person 'walk through the house' mode and a Matterport-style dollhouse cutaway, with live public demos of real rooms online. | Public | Experimental | 48% | Consumer payoff, built on MIT open-source kept as a thin fork; walk mode near-done, the differentiated dollhouse mode is the remaining custom work. |
| ARKit pose-less fast-ingest pathBecause the iPhone already tracks its own position, the pipeline skips the fragile camera-pose step — eliminating the biggest cause of failed reconstructions. | Infra | Experimental | 45% | Directly attacks the reliability problem in DIY 3D capture — the difference between 'works sometimes' and 'works for a non-expert every time'. |
| In-app tour delivery (status + push + embedded viewer)The end of the loop in the phone app: the agent sees processing status, gets a push when the tour is ready, and opens the walkable tour in-app. | Internal | Experimental | 42% | Async-by-design UX that makes a 10-40 minute cloud job feel like a delightful surprise rather than a delay. |
| Capture training-data flywheelInfrastructure to (with consent) turn every scan into a future training example, so reconstruction and guidance quality compound as more agents capture. | Infra | Planned | 15% | The compounding-moat thesis — proprietary capture telemetry rivals on commodity open-source can't replicate; currently only DB scaffolding exists. |
Pillar 10
aquienpz Asset & CDN Platform
83%
maturity · 15 capabilities
A production, multi-tenant image and video CDN platform built in-house on Cloudflare R2 + edge Workers plus a processing service — handling secure upload, automatic optimization, on-the-fly image transforms, adaptive video streaming, and a growing suite of AI photo-editing tools. It already runs the entire Realty ONE Group CR product as its reference customer (the legacy v1 asset system was fully retired in its favor) and ships published SDKs, per-tenant isolation and usage metering — the ingredients to resell as a standalone Cloudinary/ImageKit-style product. Core delivery and pipeline are mature; AI-editing, automated video composition and a customer-facing billing product are the areas still maturing.
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| On-the-fly Image CDN & Transform EngineA global image delivery network that resizes, crops and reformats any photo instantly from its URL — the same photo served in dozens of sizes without pre-generating them. | Infra | Live | 95% | Cloudflare-edge delivery means near-zero marginal serving cost and global low latency — a Cloudinary/ImageKit-style product built in-house. |
| Authoritative Image Optimization & WebP Variant LadderEvery uploaded photo is automatically re-encoded into a ladder of optimized WebP sizes (thumbnail through 4K) for fast page loads on any device. | Infra | Live | 95% | Best-in-class byte savings out of the box; the quality/size tuning is IP agencies pay SaaS vendors for. |
| Secure Direct-to-R2 Upload (BFF + Presigned URLs)Lets users upload large photos/videos straight to storage from the browser without exposing secret keys, with automatic de-duplication of identical files. | B2B | Live | 95% | A cleaner security model than Cloudinary/Uploadcare (browser never holds a key) — a real differentiator for security-conscious clients. |
| Decoupled Asset Pipeline into Neo (BFF + event bus)The integration wiring that lets the main app upload, process and display media through aquienpz seamlessly and resumably, even if the browser closes mid-upload. | Infra | Live | 95% | Proves the platform is battle-tested inside a real product at scale before being sold externally — the reference customer is the company itself. |
| Adaptive HLS Video Streaming (pipeline)Turns any uploaded video into an adaptive stream that auto-adjusts quality (240p–1080p) to the viewer's connection — Netflix-style playback for tour videos. | Public | Live | 90% | Adaptive streaming built in-house on R2 avoids per-minute Mux/Cloudflare Stream fees — a meaningful cost moat at video scale. |
| Fast Direct Video/Reel Upload PathWeb-ready videos (like marketing reels) are finalized in seconds instead of ~110 seconds by skipping unnecessary heavy processing. | Infra | Live | 90% | A measured 40x latency win on the reel path — performance engineering that drives conversion and lowers compute spend. |
| Client SDK & Upload Libraries (npm)Published developer libraries that let any app integrate image/video upload and delivery in a few lines, across web and mobile. | B2B | Live | 90% | A real, versioned SDK with a documented consumer contract is what makes this sellable to third-party developers, not just usable internally. |
| Multi-Tenant Isolation & Tenant ProvisioningServes many separate brands/agencies from one system, each with isolated storage, keys and CDN paths — the foundation for reselling as white-label. | B2B | Live | 85% | Multi-tenancy already in production is the biggest gate between 'internal tool' and 'sellable SaaS' — de-risks the productization thesis. |
| Brandless Market-Photo ProxyFor confidential 'broker mode' shares, external photos are served through a neutral, un-attributable URL that can be instantly taken down. | Infra | Live | 85% | A privacy/compliance-conscious sharing primitive that enforces controlled, revocable distribution — relevant to any regulated-content buyer. |
| AI Image UpscalingOne-click AI enhancement that increases resolution and sharpness of low-quality photos, with multiple provider options. | Internal | Beta | 80% | Pay-per-use AI provider abstraction with its own quota/pricing layer — packageable as a premium upsell tier. |
| Face-Aware Smart CroppingWhen cropping a photo to a fixed shape, the system keeps faces centered automatically instead of cutting people off. | Infra | Live | 80% | A polish feature that materially improves auto-generated thumbnails at zero user effort — a common paid add-on elsewhere. |
| Usage Metering, Quotas & Cost TrackingTracks storage, bandwidth and AI processing per tenant with per-operation cost accounting — the billing backbone for a paid product. | B2B | Beta | 75% | Metering exists but isn't a customer-facing billing product yet; converting it to invoicing/plans is a clear near-term path to revenue. |
| Dynamic Social/OG Card GenerationAutomatically generates the preview image shown when a listing is shared on WhatsApp, Facebook or iMessage. | Public | Beta | 70% | Turns every shared link into branded marketing collateral — a small but sticky feature that drives organic reach. |
| AI Photo Editing: Background Removal & Generative FillAutomatically removes photo backgrounds and fills or extends images using AI for clean product/property shots. | Internal | Beta | 65% | On-platform AI editing keeps users inside the tool instead of Photoshop/Canva — but rollout is still light, so treat as emerging. |
| AI Marketing Video Composition & SlideshowsAssembles multiple photos into marketing videos and slideshows automatically (an asset-platform path parallel to the Creative Studio). | Internal | Experimental | 55% | High-value AI feature but early-stage in this app; validate real usage before treating as shipped. |
Pillar 11
Platform, Sync & Infrastructure Foundation
87%
maturity · 12 capabilities
The technical backbone that makes the product feel instant, stay online and scale without rewrites: local-first realtime sync (back-office updates live like a Google Doc), a hardened auth and multi-tenant identity layer, self-hosted observability and CI/CD that cut recurring cloud costs, and deploy-integrity + API-scaling engineering born from real production postmortems. Most is live in production with confirmed URLs; the remaining work is reliability hardening (never-lose-a-save), not greenfield building.
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| API type-scaling architecture (Eden treaty decomposition)The type-safe frontend↔backend contract was restructured so the codebase can add hundreds more API routes without the compiler collapsing, with a CI gate blocking regressions. | Infra | Live | 95% | End-to-end type safety that survives scale is a rare engineering asset — it keeps velocity high and defect rate low as surface area explodes. |
| Local-first realtime sync (Electric SQL over SSE)The back-office reads data through a live streaming pipe, so listings, leads and boards update the moment they change anywhere — no refresh, usable on flaky connections. | Infra | Live | 93% | Local-first sync is the 'feels native/instant' moat generic CRUD competitors lack; self-hosted keeps per-seat cost near zero as agents scale. |
| Shape-level authorization gatekeeperEvery realtime stream is checked at the edge: private data needs a short-lived org-scoped signed token and the server rewrites the query so a user only receives their own org's rows. | Infra | Live | 90% | Row-level, org-scoped access control on the sync layer is the security prerequisite for safely white-labeling to multiple agencies on shared infra. |
| Authentication & multi-tenant identity (Better Auth)A production sign-in system with passkeys, two-factor, organizations/roles and admin operations, engineered to stay up (fail-open) even if a supporting cache goes down. | Infra | Live | 90% | Org/role model plus passkeys is enterprise-ready auth out of the box — required for B2B agency deals and removes a security-review blocker. |
| Self-hosted CI/CD with path-gated deploysBuilds and deploys run on the company's own runner with per-service pipelines that only trigger on relevant changes, eliminating a recurring cloud-build bill. | Infra | Live | 90% | Lower fixed infra cost plus faster, safer deploys — operational leverage that compounds as team and service count grow. |
| core-server resilience & autoscalingThe main API was re-architected after an outage: safe startup retries, cached reads, background jobs in separate processes, and at least two instances so one failure doesn't take it down. | Infra | Live | 90% | A documented postmortem-to-fix loop with quantified cost/latency outcomes demonstrates the team can run production reliably at scale. |
| Optimistic write reconciliation (txid overlay)When an agent edits something, the change appears instantly and is automatically reconciled against the database's confirmation, so the UI is both fast and eventually correct. | Infra | Live | 85% | Correct optimistic UI as a platform primitive means every future feature inherits snappy, consistent writes for free. |
| Self-hosted observability stackFull production visibility — traces, metrics, logs and error tracking — on the company's own servers, so incidents are seen in monitoring rather than reported by users. | Infra | Live | 85% | Self-hosting observability (vs Datadog per-host billing) is a durable cost advantage and signals operational maturity to acquirers. |
| Deploy-integrity guard (/health sha + stale-deploy assert)Every service reports exactly which code version it runs, and deploys fail loudly if the live version doesn't match what was just shipped — making silent stale deploys impossible. | Infra | Live | 85% | Born from a real 90-minute outage; turns deploy risk into an automated red/green gate — the reliability engineering investors look for. |
| MCP servers for AI-agent data accessDedicated servers letting AI agents query listings, leads and marketing-studio data through a standard AI-tool protocol, cleanly and with version tracking. | Infra | Live | 85% | Exposing the data layer over MCP makes the platform natively agent-ready — the foundation for AI copilots and a moat as the market goes agentic. |
| Durable collaboration streams (Yjs / CRDT broker)Multiple people and AI agents edit the same draft, blog post or board simultaneously with live presence and conflict-free merges. | Infra | Live | 80% | Real-time multi-user plus human+AI co-editing is a differentiator vs single-user real-estate tools and underpins AI-agent workflows. |
| Collaboration durability hardening ('never lose a save')A reliability program ensuring an agent editing on bad wifi never loses work: automatic reconnect, offline persistence and re-sync of edits made while disconnected. | Infra | In progress | 70% | Addresses a real data-loss incident; turning a known open-source gap into a proprietary reliability layer reduces churn and support cost. |
Pillar 12
Multi-Tenant & White-Label Productization
55%
maturity · 9 capabilities
The platform is genuinely multi-tenant at the technical level: one codebase already serves five separately-branded storefronts on their own subdomains, driven by a canonical tenant registry, and the property-render engine is packaged and running across three independent deployments (including a fully neutral, non-branded domain) — this part is live. The B2B 'sell the platform' story (EstateKit white-label + Espacio Futuro parent brand) is real in architecture but early in execution: a re-skinnable EstateKit shell is deployed as a demo but still renders fixture data, its real-backend adapter is unbuilt, and all live tenants belong to a single org. Net: strong portable foundation and a costed roadmap, but not yet sold to an outside agency.
| Capability | Audience | Status | Progress | Why it matters |
|---|---|---|---|---|
| Multi-tenant storefront engine (subdomain-per-tenant)One codebase serves several branded consumer websites, each on its own web address, switching branding and content automatically by domain. | Public | Live | 88% | Proof the platform is multi-brand by design; caveat — all live tenants today belong to one org, so it's 'multi-tenant capable, single-customer in practice'. |
| Brand-neutral co-broke share app (2cloudapp.com)A live, single-purpose brandless app that renders a shared market property on a neutral domain with zero Realty ONE branding — a co-broke share link, not a full storefront. | Public | Live | 88% | Real proof the render stack is portable and brand-agnostic — but it is a single-route share app on Vercel, not yet a standalone white-label storefront; the storefront-grade product is still EstateKit (a demo shell on fixture data). |
| Canonical tenant registryA single source-of-truth mapping each brand to its org, team, country, URL and domain so every part of the system resolves 'which tenant is this' consistently. | Infra | Live | 85% | Plumbing that makes onboarding a new brand a config change, not a code fork; schema already supports distinct orgs when a real external customer arrives. |
| Packaged, reusable property-render engineThe property-listing UI bundled as a shared library that multiple separate apps already consume, so the visual product can drop into any new site or brand. | Infra | Live | 80% | The real moat under the white-label story: the property experience already runs in three independent deployments — this is what makes 'ship a new site' fast. |
| EstateKit white-label storefront shellA standalone, re-skinnable real-estate website product — the 'sell the platform' storefront — deployed as a live demo with its own theming, multi-currency, multi-language and property detail. | B2B | Beta | 42% | The clearest embodiment of the B2B story, but it currently renders demo/fixture data — a polished shell, not yet a saleable product wired to live inventory. |
| White-label theming & site configPer-brand configuration for name, tagline, logo, colors and contact details, injected as design tokens so a new agency's storefront restyles without touching code. | B2B | Experimental | 40% | The lever that turns 'one build' into 'N brands'; currently proves the concept with a single demo config, not yet populated with real customers. |
| Pluggable data-provider abstractionA clean seam letting the white-label storefront swap between demo data and any real property backend without changing the pages. | Infra | In progress | 35% | The right decoupling to make the product backend-agnostic, but only the interface + fake data exist — the biggest thing between demo shell and sellable product. |
| Espacio Futuro parent-brand siteThe marketing website for the parent holding brand that would sit above the individual real-estate products — the corporate face of the 'we build platforms' story. | Public | Experimental | 25% | The intended umbrella brand for productizing the platform, but currently an early scaffold/landing page — signals direction more than a shipped asset. |
| EstateKit → flagship data parity migrationThe planned work to make the white-label shell render the exact same live homepage as the flagship (real listings, hero media, regions). | B2B | Planned | 15% | A detailed, costed (~1-2 week) roadmap from 'demo shell' to 'real product', but execution hasn't begun — a clear next step, not a shipped capability. |
Now / shipped
Phase 0
The live, revenue-generating platform in daily production use
- Multi-tenant storefront live in 3 languages / 4 currencies with conversational AI + map search
- Mia AI lead concierge with lead capture, routing, follow-up and Instagram auto-reply in production
- Full back-office CRM + real-time collaborative workspace used by an office of ~50+ agents
- Complete sales-closings & commission-accounting engine: closing wizard, split math, payment lifecycle, Top-Producers leaderboards and office→franchise fee reporting
- $0-marginal-cost AI video/creative engine (Takumi) producing branded social reels
- Autonomous organic marketing publishing daily across 3 Facebook + 3 Instagram brand accounts (owner-supervised)
- Proprietary market-intelligence dataset — 16 portals ingested, ~56K deduplicated for-sale listings (~59K total index docs) — with pricing heatmaps and a public market report
- In-house aquienpz asset/CDN platform + local-first Electric SQL sync + Better Auth + postmortem-hardened deploy/observability infra
Investor valueThe core business is already operating and generating value at near-zero marginal cost — this is a shipped product with real usage, not a pitch deck. The moat (owned data, owned low-cost infra, AI-native workflows) is live today.
Next 1-2 quarters
Phase 1
Operational throttle-flips that unlock revenue from already-built systems
- Connect the live WhatsApp Business channel (code-complete) — the dominant Costa Rica channel and the funnel's economic core
- Flip the follow-up/nurture consent gate and template sign-off to go-live
- Enable the governed Meta paid-ads rail and the boost-winners loop (built and verified, currently spending $0 pending owner budget OK)
- Mature the consumer identity layer (social login) and smarter alert matching to lift re-engagement
- Convert aquienpz usage metering into customer-facing billing/plans
Investor valueThe highest-value near-term gains require operational decisions, not engineering — de-risked upside. Each flip converts a sunk build cost into measurable lead volume, ad efficiency, or new revenue.
6-12 months
Phase 2
Productize the platform for external customers and let data-dependent features ripen
- Ship the EstateKit real-data adapter (data-provider abstraction) and flagship data-parity migration — turning the demo shell into a sellable white-label product
- Market Pulse lifecycle analytics (DOM, absorption, MOI) reach full signal as ~60-90 days of sweep history accrue
- Clear TikTok + YouTube Shorts platform audits to expand autonomous publishing beyond FB/IG
- Ship the real-estate voice vertical of Mia by copying the proven restaurant pattern
- Advance the Listings MCP server so inventory appears inside ChatGPT-style AI apps
- Mature generative AI video and AI photo-editing tiers as premium upsells
Investor valueOpens the second business line (sell-the-platform B2B SaaS) on top of the operating brokerage, and matures features whose value compounds automatically as data and platform audits accrue — expanding TAM beyond a single office.
Horizon / frontier
Phase 3
Frontier bets that could re-rate the platform if proven at scale
- Take iPhone-based 3D Gaussian-splat property tours from R&D (TestFlight, live demos) to a reliable, GA capture-to-walkable-tour product
- Ship native voice clients (mobile + desktop wake-word 'Ey Mia') and expand the multi-vertical voice platform across industries
- Activate the capture and agent-feedback training-data flywheels so reconstruction, guidance and marketing quality compound proprietarily
- Onboard the first external agency/franchise onto the white-label stack and the Espacio Futuro parent-brand productization
Investor valueOptionality: each is a large, differentiated market (Matterport-class 3D, catalog voice concierge, white-label real-estate SaaS) already architected on owned, commercially-clean foundations — honest R&D today, but low incremental cost to prove given the reusable engine.