QuarryFrom Fillmore Partners
The whole industry runs the same sourcing stack.We rethought the workflow.
Lower-middle-market sourcing has become a numbers game — the same databases, the same warmed-up domains, the same templated blast. Quarry doesn't run that race faster. It gives a handful of operators the research depth that used to take an army of analysts — to find the right businesses, understand them before the first call, and reach owners who feel genuinely understood.
Quarry is AI-native — intelligence is the substrate the product is built on, not a feature bolted to the side.
Allegheny Backflow & Cross-Connection
BackflowWhy now · Owner is 67 · withdrawn S-1 in 2023 · no announced sale
Ask · Confirm recurring test-contract mix; who runs the field crews
Cascade Industrial Hygiene Partners
Ind. HygieneWhy now · 3 OSHA contracts renewed · two senior PEs licensed 30+ yrs
Ask · Succession plan; share of revenue from monitoring vs. one-time
Tri-State UST Compliance Co.
USTWhy now · State registry shows 1,400 tanks under contract · founded 1991
Ask · Revenue per tank; is the founder still operating day-to-day
Illustrative. The daily surface is a prioritized action queue — the owners worth a credible touch today, the why-now, and what to ask — not a database to browse.
The problem
Outbound is commoditized. The owner deletes the email.
Anyone with a website, warmed-up domains, a CRM, Grata, Clay, and a sequencer has a sourcing machine spamming potential sellers. The standard motion competes on volume of outreach × template polish. It is shallow by construction — and owners can smell it from the subject line.
The constraints that forced firms into that motion — time, people, money, technology — have collapsed with AI. The bar is no longer do the tired playbook ten times faster. It's to do something the tired playbook can't.
The touch the whole market sends
"Hi Bob — we've been tracking your company and think there could be a great fit…"
The uncomfortable math
More outreach isn't the answer.
The standard playbook is a volume game: contact a thousand-plus owners cold, and submit almost no offers in the first year. Messaging volume doesn't manufacture conviction — and AI has only made the cold channel noisier, lowering the barrier for every buyer at once.
The deals that actually close come through relationships, depth, and a first touch that earns the conversation. Quarry is built for that race, not the other one.
The standard outbound funnel
IllustrativeA thousand-plus owners contacted cold
A handful open the email
Fewer still reply at all
A thin slice take a first call
Almost no offers submitted in year one
More at the top doesn't widen the bottom. Volume isn't the constraint — conviction is.
The contrast
The standard stack, and what Quarry does instead.
Crowded firms compete on the volume of outreach. Quarry competes on research depth per target, signal precision, and seller-side credibility on the first touch.
Crowd-sourced databases — Grata, Clay, Apollo, ZoomInfo. Thin, stale LMM coverage everyone else also rents.
Provenance-tracked primary sources — SOS filings, permits, regulator records, court filings, trade rosters. The data competitors don't read.
A universe defined by firmographics — industry code, headcount, revenue band. A list anyone can pull.
A universe defined by triggers — succession clocks, capital-cycle timing, regulatory pressure, a process that died quietly. A quieter, higher-fit list.
Warmed-up domains and a sequencer firing thousands of templated touches. Volume × template polish.
One credible touch per owner, built from a dossier deep enough that the owner thinks: these people actually understand my business.
A CRM that records what happened, then forgets why. The edge resets the day a competitor buys the same subscription.
A compounding institutional brain — every conversation, dead deal, and outcome sharpens the next thesis and the next touch.
Where the edge actually is
Sourcing has three stages. They are not equally valuable.
Identification and the offer are commodity — worth doing well, not worth competing on. The middle stage is the only one that builds something a competitor can't buy.
Identification
Build the candidate set. LMM data is broad-and-imperfect for everyone — the crowd does this part well enough.
Data collection
Turn names into a real picture of a business — owner, intent, mix, the why-now. The big firms brute-force this with armies of analysts over years. It is the only stage that manufactures a proprietary, compounding asset.
Offer & LOI
Valuation, diligence, the LOI. We build it in — financial analysis, market work, CIM, the deal record — because it belongs in one system.
Our thesis, sharpened: give four operators the Stage-2 leverage that used to require forty analysts — and do it better than the analysts, not just cheaper. Small buyers fail this stage today because they hand interns bad names and stale firmographics. Quarry wins by being devastatingly good at the first credible touch — and by capturing what every conversation reveals as the proprietary asset that sharpens the next one.
The four pillars
Depth per target, not volume of outreach.
Thesis depth at scale
Continuously generated, stress-tested sub-sector theses — grounded in primary sources the crowd doesn't read. A live thesis portfolio, not a one-off memo that goes stale in a drawer.
Signal-first targeting
A universe defined by triggers — where money has to move — not by stale firmographics. We stack signal on size, mix, geography, and ownership so the list is short and the fit is real.
Owner-specific outreach
Every approach is built from a dossier deep enough to earn the first conversation. One credible touch — never a template, never a mass sequence, never a deleted email.
Closed-loop memory
Every signal, conversation, dead deal, and outcome feeds back. The edge compounds — it doesn't depreciate the day a competitor buys the same data subscription.
Signal timeline · one owner
- 2019
Founded — single owner-operator
- Q2 ’23
Withdrawn S-1 (process died quietly)
- Q4 ’24
Two senior licenses approaching renewal
- Now
Owner age 67 · no announced succession
High-intent owner · low-competition window. A trigger the firmographic crowd never sees.
The north star
A compounding institutional brain — not another tool.
When the cost to build a feature collapses to nothing, no feature is a moat — the next firm builds it tomorrow. So Quarry isn't a toolset. It's the thing that makes the next thesis, the next touch, and the next deal smarter because of every one that came before. The research depth becomes switching cost; the captured judgment becomes an asset that doesn't reset.
The loop that closes
Every dead deal teaches more than every close. Most firms lose that lesson to a dead email thread. Quarry captures it — structurally — so the edge compounds while a fresh competitor starts at zero.
AI-native, by design
Intelligence is the substrate, not a feature.
Most software bolts a chat box onto a database and calls it AI. Quarry is built the other way around — reasoning runs through the whole product, including the parts most teams never automate. Two examples of what that looks like in practice.
Help that doesn't hallucinate
A corpus-grounded help assistant
Ask Quarry anything and it answers from Quarry's own documentation — and cites the sources it used. When the answer isn't in the docs, it says so plainly instead of inventing one. Honest help, not a confident guess.
How does auto-archive decide what to drop?
Only a verified, order-of-magnitude buy-box miss auto-archives — and it's always reversible from the Briefing. Inferred or edge cases stay a recommendation for you to review.
Off-corpus · I don't have that in the docs, so I won't guess. Here's who to ask instead.
Product management that runs itself
Automated feature-request triage
Submit a request in-product and an AI product manager restates it, scopes it against the firm's mandate, dedupes it against everything already open, t-shirt-sizes it, and recommends. The backlog triages itself.
- 1
Restate
in the firm's own terms
- 2
Scope
against the mandate
- 3
Dedupe
vs. open requests
- 4
Size
t-shirt estimate
Ship · S · merges with an open request
And the product tests itself
A synthetic-user harness drives the real app end-to-end — AI personas that act like our operators, grade the experience, and surface friction before a human ever hits it. The product is continuously exercised by the same kind of intelligence it's built from.
Built by operators, for operators
We built this to source our own platform.
Quarry was built by Fillmore Partners — a permanent-capital holding company — for its own direct-acquisition program. We aren't a vendor selling a sourcing tool; we're acquirers who needed the firm's knowledge to compound instead of evaporate. The hard part of sourcing was never a longer target list. It was converting names into a real picture of a business, earning a credible first conversation, and never losing what we learned. Quarry is how we do that — and now a small group of acquirers run on it too.
Proof, not promises
This is what AI-native building looks like.
Quarry is a real, multi-tenant sourcing platform — not a demo, not a vibe-coded toy. A small team built it fast, end to end, using the same AI-native methods the product runs on. The grounded help assistant, the self-triaging feature pipeline, and the synthetic-user harness aren't marketing — they're how the thing was built and how it keeps improving. When the cost to build collapses, the firms that win are the ones who actually build. We did.
Grounded help, shipped
A help assistant that answers from the docs and declines when it can't — live in the product.
A backlog that triages itself
Requests restated, scoped, deduped, and sized by an AI PM — product management as infrastructure.
Tested by synthetic users
AI personas exercise the real app end-to-end and grade it, surfacing friction before users do.
Request access
Quarry is in a closed beta with a small group of acquirers — searchers, serial consolidators, and the investors behind them. Tell us a little about your mandate and we'll be in touch.