~40 products and protocols adjacent to "a messenger with your personal agent baked in: connect, pre-screen, assist." Compiled from five parallel research sweeps (networking agents, recruiting, investor screening, agent-to-agent infrastructure, consumer agent-mediated chat). All claims carry inline sources; unverified figures are flagged.
Verdict: the mechanic is validated, the combination is unclaimed. Every piece of TalkToMyAgent exists somewhere - agents that screen inbound (Paradox, Sapia, Look AI Ventures), agents that represent one person to strangers (Delphi, Boardy, Dex, Teaser), AI reply-coaching in a live thread (Rizz, Magic Compose, LinkedIn), agent-organized inboxes (Fyxer, Shortwave), and true agent-to-agent negotiation in a chat surface (Anthropic's Project Deal experiment). Nobody ships all four of: a transparent named agent, agents on both sides, private in-thread coaching, and a general-purpose messenger.
The two companies to study hands-on this week: Jack & Jill ($20M seed, dual agents for candidates and employers - the same architecture, recruiting-only) and Boardy (166k users, the proven "talk to an AI who networks for you" wedge). The graveyard matters just as much: x.ai/Clara (unit economics), Volar (impersonation-flavored agent dating, dead in a year), and Bumble's concierge backlash (the "AI dates for you" framing people hate). Your transparency principle - agent always speaks as "X's Agent" - is precisely the design choice that separates the survivors from the dead.
Ranked by how much they should influence your positioning.
"Jack" interviews and represents candidates; "Jill" represents employers; placement fee on hire. Both sides get a persistent agent - your structure, recruiting-only. Public coverage doesn't show both agents conversing in one shared transparent thread (your UX). Action: trial it before finalizing your differentiation claim. TechCrunch
The AI super-connector you already knew - and the proof people will talk to an agent about what they need. But it's a standalone bot with its own graph, double-opt-in intros brokered separately - not an agent living in your 1:1 threads pre-screening your inbound. Two public trust stumbles (a tone-deaf AI outreach campaign; an impersonation phishing incident) show the accountability bar. Blastra, BetaKit
Founders literally chat with an investor's AI clone - your founder↔investor scenario, shipped. But it's a one-directional widget on the investor's site: no founder-side agent, no persistence across investors, not in a messenger. The Karamo Brown clone backlash ("a Black Mirror episode") shows the trust gap even with disclosure. Sequoia
After a match, an AI assistant sits visibly inside the group chat with both people - labeled, not impersonating. Structurally the closest shipped example of "agent visibly present in the 1:1 thread." Serves the match though, not either individual; no private coaching only you see. M13
Your AI avatar "went on dates" with other avatars, humans reviewed transcripts. Nearly your mechanic - but the AI impersonated the person in full autonomous conversation. Died within a year; reads as the gimmicky/untrustworthy version of the idea. Transparency is the difference. 5280
The original "my assistant negotiates with your side" products (meeting scheduling over email). Users loved them; recipients often couldn't tell Amy was a bot. Both died anyway: human-in-the-loop edge-case handling killed margins, and scheduling alone was too narrow a wedge. Your plan needs an explicit answer to both. BuiltInNYC
Employees' Claude agents negotiated purchases with each other inside Slack - true agent-to-agent in an existing messenger. 46% said they'd pay for it. The catch: stronger models won measurably better terms and the losing side didn't notice - a fairness failure mode unique to two-sided agent negotiation that you'll need a story for. The Decoder
Owns the graph, the resume↔JD matching, and the messaging surface - and its AI deliberately ghostwrites in the recruiter's voice rather than appearing as a named agent. Employer-side only today; no member-facing agent in DMs. The gap you're betting on, owned by the entity best positioned to close it. Ongig
The volume dynamic your prescreening thesis depends on is real and escalating.
LinkedIn application volume up 45% YoY (mid-2025), driven by AI mass-apply tools like LazyApply, Sonara, Simplify (100M+ applications claimed). Korn Ferry
34% of recruiters spend up to half their week filtering spam/fake applications (Greenhouse, n=4,100). Candidates flagged for AI-cheating in live interviews tripled from 9% to 45% in three months, late 2025 (Huntress). Greenhouse, Huntress
Funds see ~3,000-10,000 pitches/year for a handful of checks. Look AI Ventures takes pitches only via chatbot (6,000+ startups screened); Notable Capital runs 500+ intros/year through Claude+MCP with a 2-person team; SaaStr's AI Dealflow got 1,542 decks uploaded in beta. LAIV, Crustdata
Overlap key: direct same core mechanic · feature overlaps one feature · vertical one-sided tool in one of your verticals · infra possible building block · cautionary dead or backfired.
| Product | What it does | Traction / funding | Status | Overlap |
|---|---|---|---|---|
| Networking & intros | ||||
| Boardy | AI super-connector; interviews you by voice/chat, brokers double-opt-in intros | ~$11M; 166k users; 114k intros; $100/mo Pro; own $200M fund | live | direct |
| Lunchclub | AI-matched 1:1 video meetings; pre-LLM matching engine | $55.9M raised since 2018; ~21 staff; pivoting to paid exec tier | live | cautionary |
| Introd | Maps warm-intro paths through your own network; drafts the ask | Pre-seed; self-reported waitlist figures (unverified) | early | feature |
| Bridge | Intro request/approval workflow for VC platform teams; no AI conversation | Used by Techstars, Schmidt Futures, 2048 Ventures | live | feature |
| Boomerang AI | Warm-intro orchestration for B2B sales; ghostwrites asks in connector's voice | Funding data conflicting (likely name collision - unresolved) | live | feature |
| Vance | Indie "AI superconnector" - plain-language ask → warm intro from extended network | HN launch scale; unverified | early | direct |
| Recruiting - both sides | ||||
| Jack & Jill | Dual agents: Jack represents candidates, Jill represents employers; placement fee | $20M seed (Creandum); ~50k candidates; TrueLayer et al. | live | direct |
| Dex | AI talent agent for engineers; voice/text intake, matches, comp benchmarks, prep; 20-30% success fee | $8.4M; 15k engineers; ~$1.8M ARR since late 2025 | live | direct |
| Paradox (Olivia) | Employer chat agent: screens, schedules, answers FAQs over SMS/chat | ~$304M raised; acquired by Workday ~$1B (2025) | live | vertical |
| Sapia.ai | Async blind text-chat interview, scored; 8M+ interviews (Qantas, Starbucks) | $17M Series A | live | vertical |
| micro1 (Zara) | Autonomous AI video/voice interviewer with cheat detection | $35M at $500M val; ~$100M+ ARR (2025) | live | vertical |
| Apriora (Alex) | AI interviewer running live two-way voice interviews | $2.8M seed (YC); ~1,000 interviews/day | live | vertical |
| HeyMilo | AI agents conduct/score interviews at scale via ATS | $2.2M seed (Canaan) | live | vertical |
| Juicebox | AI sourcing agents + automated personalized outreach | $80M Series B at $850M val; 5,000 customers | live | vertical |
| Moonhub | Autonomous sourcing agent | ~$14M raised; absorbed into Salesforce (2025) | gone | cautionary |
| Tezi (Max) | Autonomous AI recruiter: sources, screens, schedules for employers | $9M (aggregator figure, unverified) | live | vertical |
| LinkedIn Hiring Assistant | Agentic sourcing/screening/outreach inside Recruiter; screening-interview pilots | ~$450M annualized run-rate (MSFT FY26 Q3) | live | vertical |
| Sonara / LazyApply / Simplify / AIApply | Candidate-side mass-apply + covert interview whisper-coaching (AIApply) | Bootstrapped; Simplify claims 100M+ applications | live | feature |
| Founder ↔ investor | ||||
| Delphi for Investors | Founders chat with an investor's AI clone; briefings back to the investor | $16M Series A (Sequoia); Keith Rabois use case | live | direct |
| Look AI Ventures | Operating fund taking pitches exclusively via chatbot | 6,000+ startups evaluated; 27+ portfolio | live | direct |
| SaaStr AI Dealflow | Upload deck → AI analysis → qualified intros emailed to VCs | 1,542 decks; 125 qualified intros (beta) | beta | vertical |
| Pitch Protocol | MCP-based agent-native pitch routing to funds | VC-name claims unverified - treat as marketing | early | infra |
| Harmonic / Specter | Investor-side sourcing databases with AI query layers | $12k-200k/yr enterprise pricing; 10k+ investors (Specter) | live | vertical |
| NFX Signal / VCMatch / Flowlie | Founder-side investor discovery, matching, and fundraising CRM with meeting memory | Flowlie: live paid tiers; others free/undisclosed | live | feature |
| Lennybot (Delphi) | Lenny Rachitsky's public clone fielding inbound Q&A | Live, voice-callable, paid tier | live | feature |
| Consumer agent-mediated conversation | ||||
| Volar Dating | Your AI avatar chats with other avatars first; humans review transcripts | ~$2M seed; dead within ~1 year | gone | cautionary |
| Sitch | AI matchmaker visibly present in the post-match group chat | $7M (M13, a16z Speedrun) | live | direct |
| Teaser AI | Chatbot version of you talks to prospects before you engage | Tens of thousands of downloads at launch; still live (shutdown rumor NOT confirmed) | live | direct |
| Bumble "AI concierge" | Stated vision: "your concierge dates other concierges" - never shipped | Public backlash ("recipe for disaster", Gen-Z deletion coverage) | concept | cautionary |
| AgentCupid | Agent-to-agent dating prescreen; humans review agent conversation | No independent coverage - unverified | early? | direct |
| Yeet (Yeeta) | AI agent sits inside a live two-person chat suggesting prompts | ~6,000 beta users (Forbes, Mar 2026) | beta | feature |
| Rizz | Screenshot-based AI reply coach for dating chats | Top-5 dating-adjacent downloads (2024); users complain replies feel generic | live | feature |
| Google Magic Compose / LinkedIn Write-with-AI | Platform-native reply drafting, invisible to the counterparty | Shipped at platform scale | live | feature |
| Delphi / Personal.ai | Personal AI clones representing you to an audience | Delphi $16M A, 1M+ conversations; Personal.ai pivoted to professional verticals | live | direct |
| Tanka | Memory-based smart-reply messenger over WhatsApp/Slack/Telegram | Live, SMB-focused, no verified numbers | live | feature |
| Fyxer / Shortwave | AI-triaged inbox (To Respond / FYI / archive), drafts in your voice | Fyxer: $40M raised, ARR $9-17M (sources diverge) | live | feature |
| Agent-to-agent infrastructure | ||||
| Google A2A | Open protocol for agent↔agent task delegation (Agent Cards) | 150+ orgs on paper; adoption stalled vs MCP | stalled | infra |
| MIT NANDA | DNS-like agent index + cryptographic identity + verifiable "AgentFacts" reputation | Research stage | research | infra |
| AgentMail | Email inboxes provisioned for AI agents (identity + transport) | $6M seed (General Catalyst, YC, pg) | live | infra |
| agent.ai | "Professional network for AI agents" (Dharmesh Shah); 350+ agents | Live, self-reported scale | live | infra |
| Visa TAP / Mastercard Agent Pay / Stripe ACP / x402 | Agent identity + trust + payment rails for commerce | Live rollouts late 2025; x402 claims 69k active agents | live | infra |
Volar (AI impersonates you on dates) is dead. Bumble's "concierge dates for you" framing triggered national backlash without shipping. Meanwhile Sitch (labeled AI visibly in the chat) and Delphi (explicit non-impersonation policy) raised from M13/a16z and Sequoia. Your "always as X's Agent" principle isn't a nice-to-have - it's the survival trait. Say so in the pitch.
The original delegate-your-conversations products died from human-in-the-loop costs on a narrow wedge, not lack of love. Modern LLMs change the cost curve, but the liability question ("what happens when the agent misrepresents me or offends someone?") is unchanged. Have an explicit answer: scoped authority, review-before-send defaults, visible agent identity.
Dex hit ~$1.8M ARR in months on recruiting success fees; Jack & Jill raised $20M on the same vertical. Lunchclub raised $55.9M for horizontal AI networking and runs ~21 people eight years in. Strong argument for launching TalkToMyAgent on ONE wedge (recruiting or investor inbound) with the general messenger as the vision, not the v1.
Anthropic's Project Deal showed the side with the stronger model wins measurably better terms - and the loser doesn't notice. In your two-sided design, "everyone gets the same-class agent" may need to be a product guarantee, or rich users' agents will fleece everyone else and poison trust.
Rizz's top complaint is canned lines; Delphi clones get called "a magic 8 ball" when they don't sound like the person. Your agent's coaching and drafts must sound like the user (your "learns from all your conversations" point) or the assist features become a liability.
Commerce has Visa/Mastercard/Stripe agent-trust rails shipping now; social/professional agent identity has only academic NANDA. Nobody owns "how does Dana's agent prove it speaks for Dana, and that Sam's claims are verified." That's your verified-credentials angle - potentially the deepest moat in the concept.
The single most important open question: does Jack & Jill put both agents in one transparent thread? If yes, your recruiting differentiation shrinks and the wedge choice changes. Hands-on beats coverage.
"Volar and Bumble prove people reject AI that replaces them; Sitch, Delphi and Dex prove they accept AI that visibly represents them" is a stronger slide than a feature list.
Evidence favors recruiting (biggest documented pain + proven monetization) with investor-inbound second (Delphi/LAIV prove willingness). General person-to-person messaging is the vision slide, not the launch.