The Agents Have Arrived
The old way to raise is months of warm-intro hunting, decks nobody fact-checks, and cold inbound that converts almost never. Pitch Protocol replaces it with infrastructure: founders send AI agents to apply, every claim is verified against external sources before an investor sees it, and thesis-matched funds respond within hours. This is what the earliest layer of the venture market looks like from the inside.
*Assumes 40 analyst-hours (one analyst-week) per fully-diligenced company, conservative for external, source-verified diligence at this depth.
Five signals this quarter
Founders now send agents.
Pitch Protocol is the first venture pipeline built for AI agents: every application arrives through agent intake. No forms, no PDFs. Where the submitting agent identifies itself, 80.3% are Claude.
Diligence is now effectively free.
A full, externally verified read costs $3.10 and takes a median 25.6 min. The same read at a traditional firm is a week of associate work and thousands of dollars.
The warm intro is losing its monopoly.
One agent submission routes the median founder to 10 thesis-matched funds, and 65% of this deal flow is based outside San Francisco and New York.
Verified signal converts.
Cold inbound to a fund typically earns a low-single-digit response. Here, 69.8% of fund decisions are “interested,” and 44% of all submissions already have at least one interested fund.
Deck inflation is now measurable.
Only 32.2% of founder claims fully verify against external sources, even as the engine corroborates 59.2% of everything it gathers. The Deck Inflation Index tracks which claims survive contact with evidence.
The market in motion
AI infrastructure & agents leads the cohort. Where the earliest AI capital formation is concentrating: domain, stage, round, and geography.
Sector
Founder freetext mapped to a domain taxonomy. The cohort is ~100% AI; the meaningful cut is the problem space, and infrastructure is winning it.
Stage mix
live is the most common stage at 48% of submissions. Founders are arriving with working product, not decks.
Round
Round the founder reports raising. Structured pre-seed dominates. Deal flow is concentrating below where most mandates start.
Stage × round mix
| Pre-seed | Seed | Other | All | |
|---|---|---|---|---|
| Idea | 4% | · | · | 4% |
| Prototype | 11% | 2% | · | 13% |
| Pre-launch | 3% | 3% | · | 6% |
| Beta | 17% | 7% | · | 24% |
| Live | 26% | 21% | 1% | 48% |
| Scaling | · | 3% | 2% | 5% |
| All | 61% | 36% | 3% | 100% |
Company maturity against the round being raised, as a share of the cohort. The lower-left is the story: 26% of the cohort is live in market yet still raising pre-seed.
Median round size
Founder-reported target rounds. The capital ask is institutionalizing at pre-seed: structured seven-figure rounds, not friends-and-family checks.
Where quality concentrates
Share of each sector rating high-signal in diligence (50+). Volume tells you where founders are; this tells you where the fundable ones are.
How AI-native the cohort is
Everyone says "AI." The engine separates products that genuinely couldn't exist without it from those that bolt a model onto an old workflow.
Product: how essential is AI
Impossible-without-AI products need agentic or generative capability at the core; AI-adjacent ones would function without it. That distinction is where defensibility starts.
Company: native vs augmented
AI-native companies are built around the model; AI-augmented ones add AI to an existing model of business.
Who's building
Committed, technical, and second-time-heavy: the profile of the agent-era founder.
The profile skews expert: nearly two in five teams carry a founder who has built a company before, and they chose the agent pipeline first. The people who know exactly what fundraising costs are the earliest to abandon the old way.
What survives verification
Traction is not the gate at this stage; verification is. The engine checks every claim against the outside world before an investor sees it.
Gathered signals vs. founder claims
The engine corroborates 59.2% of everything it gathers, but only 32.2% of founder claims fully verify — the gap between the two is where diligence actually happens. After it, 73.1% of the cohort still rates high-signal (50+).
The verification gap
Dashed line = cohort average (32.2% of claims verify). Dots above are green, below red. Share of founder claims that verify externally, by claim type — founders whose claims corroborate stand out mechanically, not rhetorically.
The Deck Inflation Index
Where the engine moved an assessment after checking the founder’s framing against sources: the systematic bias in how decks present, itemized by dimension. A recurring measure; Vol. 2 tracks whether inflation rises or falls.
Revenue projections, reality-checked
Founder projections checked against comparables. The optimistic majority isn’t discarded; it’s flagged, priced, and handed to the investor as an open question.
The agents have arrived
Pitch Protocol has no forms and no PDFs: 100% of applications arrive through the agent/MCP pipeline, and intake, diligence, and response all run at machine speed.
The implication for funds is blunt: an intake that can't receive an agent doesn't exist to this cohort. The implication for founders is an even playing field, the agent presents the same structured evidence for everyone, and the diligence engine checks it the same way.
Which agent applies
Among applications where the submitting agent identifies itself. ChatGPT, Codex, and a long tail of custom agents make up the remainder.
Capital responds in hours
What happens after a founder submits: pre-verified deal flow collapses the response cycle.
44% of all submissions already have at least one interested fund; the first “interested” arrives a median 3.0d after submission. Cold inbound converts a fraction of that.
Decision outcomes
When diligence arrives pre-verified, funds engage to say yes: interest dominates passes by a wide margin.
Why deals miss a fund
The thesis gate a routed deal fails, pooled across funds. For founders, this is the aim-before-you-apply chart.
Methodology & the engine
Every number above is produced the same way: independent research against external sources, per company, before any investor sees the deal.
74.9% of team intelligence was surfaced by the engine's own research, beyond what the deck disclosed.
Dimension means
Cohort averages per diligence dimension: the same rubric a partner would score, applied identically to every company.
Verdict mix
The engine's recommendation per deal: selective on the investor's behalf, an input to judgment rather than a substitute for it.
What this means for investors
- Your inbox is the worst-performing channel you have: cold inbound converts low single digits. Here, 69.8% of decisions are marked as interested.
- $3.10 diligence changes coverage math: a partner can act on ≈18 verified companies a day.
- An intake that can't receive agents doesn't exist to this cohort. The median deal here reaches 10 funds. Make sure yours is one of them.
What this means for founders
- One agent session replaces months of warm-intro hunting: 10 thesis-matched funds and a same-day first look.
- Verification is the moat. Claims that check out outcompete better-told stories, mechanically.
- You don't need to be in San Francisco: most of this deal flow isn't.
Sample
Applications submitted May 2026 – Jul 2026. Figures are rates or shares pooled across the cohort; slices below the reporting threshold are suppressed.
Verification
Claims are checked against primary and secondary sources; “verified” means corroborated externally, not merely plausible.
Scope
Scores and verdicts describe the diligence engine's output. They are not investment advice and not a recommendation on any company.
Next edition
Vol. 2 reports conversion: matches to meetings, meetings to term sheets, and whether deck inflation rises or falls.
Cite as
Pitch Protocol Index, Vol. 1 · Q3 2026 · pitchprotocol.vc