Investor deck · October 2026 · Confidential
Proxy gives every single a personal agent that knows exactly what they want, including the parts nobody writes on a profile. The agents meet each other. The humans meet only when both agents say yes.
The problem
Every incumbent is paid for time on app. A match that works is a lost customer. So the product is tuned for hope, not outcomes, and the users have noticed.
The information that predicts a good relationship (temperament, pace, what someone needs in bed and after it) is exactly the information nobody puts in a profile. A profile is a performance. A brief to a private agent is a confession.
Sources: Forbes Health / OnePoll dating-app burnout survey via Global Dating Insights; Tinder payer figures via AOL.
The product
A 40-minute private conversation with your agent: looks, mind, temperament, dealbreakers, intimacy, timing. The agent asks the questions a matchmaker would and a profile never does.
The brief is encrypted and never rendered as text to another human. Not to a match, not to staff, not in a screenshot.
Your agent meets every other agent in your metro. Hard dealbreakers first, then weighted compatibility across the full brief. Thousands of conversations a week, while you sleep.
When both agents score above threshold and both say yes, each human receives one introduction with a plain-words note on why. Either side can decline silently.
"I met 1,204 agents this week. I said no to 1,203. Meet Dana, Thursday, 7pm. Here is why."
An introduction from a member's agent, private beta.
Four creative directions for the consumer brand are in test. See slide 9, or open them from the Product tab.
Why now
Frontier models hold a 40-minute brief with nuance, negotiate with another model, and explain a decision in plain words. Two years ago this was a demo. Today it is a product.
The swipe generation is leaving. Matchmaker interest is surging among Gen Z, and three of the four largest apps now ship AI assistants bolted onto the swipe.
Agent-to-agent dating has appeared in Hong Kong (Clawmistry) and as web tools (AgentCupid, MoltMatch). AI setup services charge $20 a date (Amata) to $160 a pack (Sitch, $6.7M raised). Demand is proven. Nobody owns the brand, the safety layer, or the full-spectrum brief.
Sources: Axios, May 2026; Global Dating Insights on Sitch; Techleap on Clawmistry.
Market
TAM
SAM
SOM
Proxy does not need to win share from swiping to work. It needs the share of singles who have already quit swiping and would pay a professional to do it properly. That segment is the fastest-growing in the category and has no scaled product.
Sources: Straits Research, 360iResearch. SAM and SOM are company estimates.
Monetization
| Line | Who pays | Price | Why it holds |
|---|---|---|---|
| Plus subscription | Active singles | $29 / mo | Unlimited negotiations, up to 3 introductions a week, the "why you" report. Priced at a premium to Bumble's $27.61 ARPPU because the product does the work. |
| Black concierge | Time-poor, high-income | $99 / mo | Verified-only pool, a human concierge, first date planned and booked. Matchmakers charge $3k–$20k for a worse version of this. |
| Pay-per-introduction | Free tier | $9 / intro | Charged only on a mutual yes. Aligns our revenue with outcomes and converts the price-sensitive without a subscription. |
| Partners | Matchmakers, events, venues | Revenue share | White-label agent matching for human matchmakers and singles events; venue partnerships for the booked first date. |
Unit economics · base case
| Assumption | Value | Note |
|---|---|---|
| Blended CAC per signup | $11 | 40% of signups organic or referred; paid CAC $18 |
| Free to paid conversion | 16% | Paywall sits at the reveal, after the agent has already found someone |
| CAC per paying member | $69 | $11 ÷ 16% |
| Blended ARPPU | $32 / mo | 80% Plus, 8% Black, 12% pay-per-intro |
| Inference cost per payer | $2.60 / mo | Small model for dealbreaker pass; frontier only for final negotiations |
| Gross margin | 76% | Inference, verification, payment fees, support |
| Average paid life | 8 months | Members graduate; we count that as a win and a referral |
| Contribution LTV | $195 | $32 × 8 × 76% |
| LTV / CAC | 2.8× | 3.4× including the alumni referral loop |
| Payback | 2.8 months | $69 ÷ $24.30 monthly contribution |
Benchmark: Bumble App ARPPU $27.61 and 2.19M paying users, Q4 2025 (company infographic). All Proxy figures are model assumptions, not results.
Projections · base case
| Y1 · 2027 | Y2 · 2028 | Y3 · 2029 | |
|---|---|---|---|
| Cumulative signups | 180k | 1.1M | 3.6M |
| Paying members, year end | 18k | 110k | 330k |
| Revenue | $2.4M | $21M | $74M |
| Gross margin | 61% | 72% | 76% |
| EBITDA | −$6.1M | −$3.8M | +$12M |
| Metros live | 3 | 12 + UK | 30 + EU |
Base case. Year 1 margin is lower while negotiation batching and the small-model dealbreaker pass are tuned. Downside and upside cases in the data room.
Go to market
Two-sided consent means the product only works where both agents live. We launch metro by metro, seeding each with partner matchmakers and singles events so the first thousand briefs arrive before the first ad. New York, then Austin and London.
Acquisition leans on earned media and creators. A dating app where nobody reads a profile, and where you tell a machine what you want in bed, is a story that writes itself. We are testing four creative directions for the consumer brand, each with its own edge:
Competition
| Incumbents Tinder, Hinge, Bumble + "Bee" | AI setup services Sitch, Amata, Joey | Agent-native Clawmistry, AgentCupid, MoltMatch | Proxy | |
|---|---|---|---|---|
| Who reads profiles | You | A matchmaker and a model | Agents | Agents |
| Intimacy in the match | No | Rarely | Thin | Full brief, sealed |
| Both sides must consent | Mutual like | One-sided setup | Varies | Both agents, then both humans |
| Revenue tied to outcome | No | Per date | No | Per introduction + tiers |
| Trust and safety layer | Mature | Human | None | ID, zero-render, agent policy |
| Consumer brand | Strong | Local | None | Four in test |
Millions of sealed negotiations train a compatibility model that no profile dataset contains, and that cannot be scraped.
Both agents must be on Proxy. Metro density compounds; a late entrant starts with an empty room.
Zero-render architecture becomes the reason to confess. Incumbents cannot retrofit it onto a profile product.
We are paid on outcomes, so retention comes from reputation and referral, not from engineered hope.
Trust, safety and regulation
A brief is encrypted at rest and only ever processed by the member's own agent inside a negotiation. No human, including staff, sees it as text. Introductions carry a summary, never the source.
Both agents must say yes; both humans confirm; either can decline with no notification. Members revoke or rewrite the brief at any time and the agent forgets.
18+ with government ID and liveness check. Agents refuse any brief that is not about consenting adults. A human trust-and-safety team reviews escalations, never briefs.
Sexual orientation and preference are special-category data under GDPR and several US state laws. Explicit consent, purpose limitation, deletion on demand and no data sale are designed in from day one. App stores: 17+/18+ rating.
| Risk | Mitigation |
|---|---|
| Inference cost | Dealbreaker pass on a small open model; frontier model only for final negotiations; batching across the metro graph. Target under $2.60 per payer per month. |
| Cold start | Metro-by-metro launch; partner matchmakers and events seed the first briefs; waitlist gated by density. |
| Brand backlash to the risqué angle | Four directions in test; the "Unsaid" story earns media while "Stop Swiping" and "Kismet" carry the mainstream. |
| Platform policy | Native apps rated 18+; web product as the primary surface; no explicit content is ever generated or stored. |
The ask
| Milestone | When |
|---|---|
| Private beta, New York, 5k briefs | Q1 2027 |
| Three metros, 50k members, pay-per-intro live | Q3 2027 |
| Black concierge, 12 metros, London | 2028 |
| 100k paying members, EU launch, Series A | H1 2029 |
$12M Seed+ on a priced round. 24 months of runway on the base case. Lead investor sought; two strategic angels from the category committed.
Founder and CEO: the person who sent you this deck. Hiring now: Head of ML (agent negotiation), Head of Trust and Safety (ex-platform), Head of Growth (consumer subscription). Advisors from matchmaking and consumer AI.
Cohort model, downside and upside cases, the agent negotiation protocol spec and the safety policy.