Evidence-first opportunity engine

Opportunities you can defend.

Paste a website. Vox Prospect reads what it sells, maps who is showing real buying intent, and returns a short stack of proof-backed opportunity packs — each one held for your approval before anyone is contacted.

Start with a website → See opportunity pack
  • 09±Packs per run, median — not thousands
  • 01+Source citation minimum, per pack
  • 00Messages sent without your approval
Pack 07 / 09 · Run VP-0142 Awaiting approval

Fernhollow Outfitters

Mara Okafor — Operations Lead · Portland, ME

Ref VP-0142-G
Opened 2026-07-01

Evidence · Verbatim
“We’re on our third inventory spreadsheet system this year and my team still can’t tell what’s actually in the warehouse. Genuinely open to anything at this point.”
SRC forum.opsleaders.example/threads/8842 CAPTURED 2026-07-01 09:14 UTC

Why fit

Multi-location retail inventory is the exact workflow your product replaces. Team size (~40) matches your best-retained accounts.

Why now

Post is six days old. Author is the decision-maker and describes an active, failed migration.

Fit87/100
Intent74/100
Confidence81/100
Risk

Single-source signal. One public post; no corroborating source yet. Treat intent as probable, not certain.

Held in queue — your call
§ 01The case against volume

Twelve thousand rows is not a pipeline.

Most prospecting tools optimize for the size of the export, because the export is the product. Nobody has to defend a row.

The standard offer works like this: pay a subscription, receive a spreadsheet. Names, guessed emails, a firmographic filter or two. No reason to believe any given row wants what you sell. No reason to write to them this week rather than never.

So the volume gets pushed downstream — onto you, and onto thousands of strangers who never signaled anything. The real cost isn’t the subscription. It’s your name, spent one ignored message at a time.

Vox Prospect starts from the opposite premise: an opportunity is only real if you can point at the sentence that proves it.

  1. I.Fewer, argued.A short stack of cases, each one built to survive your skepticism.
  2. II.Sourced, quoted.Every pack carries a verbatim quote and a live link. Check it yourself.
  3. III.Approved, then acted.Nothing reaches a human being until you have read the proof and said yes.
leads_export_final_v3.csv12,408 rows
Why this one? No source Guessed
Meanwhile, one Vox Prospect pack

“…third inventory spreadsheet system this year… genuinely open to anything.”

SRC forum.opsleaders.example/threads/8842 · Why-fit + why-now attached · awaiting your approval

A typical export

12,408 rows · 0 citations · 0 reasons

against
A Vox Prospect run

9 packs · 9 sources · 9 argued reasons

§ 02How a run works

From one URL to a queue of defensible cases.

Every run keeps a chain of custody. Each stage produces a named artifact you can open, edit, or reject — nothing is a black box between the website you paste and the action you approve.

Chain of custody Run VP-0142 · loamandledger.example
T+00:00input
01

You paste a website.

Yours, or any product you operate. That single URL is the only setup a run requires.

Produces raw source material
T+00:02analysis
02

Product profile.

What it sells, who it’s for, what it replaces, and what a good customer looks like — written in plain language and fully editable before anything else happens.

Produces product profile · editable
T+00:05analysis
03

Buyer-intent map.

The specific, observable signals that suggest someone needs this product — complaints about alternatives, visible workflow gaps, active procurement language. Not personas. Signals.

Produces intent map · reviewable
T+00:11collection
04

Source hunters.

Readers work only the public lanes you’ve enabled — forums, reviews, filings, job posts, community threads. Everything they bring back arrives with its URL and capture time attached.

Produces raw signals · cited
T+00:26synthesis
05

Opportunity packs.

Signals become argued cases: verbatim quote, source, why-fit, why-now, scores, risk. A signal that can’t carry a full argument is discarded, not padded.

Produces opportunity packs · typ. 6–12
T+00:27holding
06

Approval queue.

Packs wait here. They do not expire into automatic sends. They do not escalate. They wait for a person.

Produces a queue · nothing moves
T+youdecision
07

Approval. The human gate

You read the proof and approve, hold, or discard — and you can edit the recommended action before approving it. Your reasoning is logged with the pack.

Produces your decision · on record
Afteraction
08

Controlled next action.

The approved step, exactly as you approved it: a reply draft, an intro note, a call brief. You send it from your own accounts, on your own schedule — Vox Prospect never contacts anyone on its own.

Produces outreach you can stand behind
§ 03The core object

Anatomy of an opportunity pack.

Not a row. Not a “lead.” A pack is a small case file: the claim, the proof, the reasoning, the risk, and the one action worth taking — built so you can interrogate every part of it.

Pack 02 / 08 · Run VP-0158 Awaiting approval

Copperline Freight Co-op

Deniz Aral — Head of Fleet Operations · Asheville, NC

Ref VP-0158-B
Opened 2026-06-29

1 Evidence · Verbatim
“Our dispatch board is three whiteboards and a group chat. We quoted two route-optimization tools last spring and both wanted 12-month contracts before a pilot. Still looking.”
2 SRC reviews.fleetstack.example/copperline#r-2291 CAPTURED 2026-06-29 14:03 UTC · ARCHIVED COPY ON FILE

3 Why fit

A 28-vehicle regional fleet describing exactly the dispatch workflow your product replaces — with their main objection (contract length) already on record, which your monthly pilot directly answers.

4 Why now

A procurement attempt within the last quarter, and the author still marked “open to vendors” as of capture. The window is real but not indefinite.

5 Fit91/100
Intent82/100
Confidence77/100
6 Risk

Quote is eleven weeks old — re-verify before contact; they may have signed elsewhere. Co-op structure means the author may share the decision with a member board.

8 Held in queue — your call

Field notes — read against the markers

  1. 1
    Evidence, verbatim

    The quote is stored exactly as found — never paraphrased, never “enriched.” If the quote can’t carry the case on its own, the pack doesn’t ship.

  2. 2
    Source, live

    Every quote keeps its URL and capture time, with an archived copy preserved. Click through and check it yourself — skepticism is the intended workflow.

  3. 3
    Why fit

    A plain-language argument mapping their stated situation onto your product profile. If the mapping is a stretch, the pack says so — or doesn’t exist.

  4. 4
    Why now

    Recency, urgency, and whether the voice on record can actually decide. Timing is half of every opportunity; it’s never left implied.

  5. 5
    How scores are derived

    Fit measures overlap between their described problem and your product profile. Intent weighs the specificity and recency of the signal. Confidence is source reliability crossed with corroboration. Every number traces back to sentences — and low numbers are shown, not hidden.

  6. 6
    Risk & limitations

    What the pack doesn’t know, stated before you act: single-source flags, stale-signal warnings, ambiguity about who decides. An honest case includes its weaknesses.

  7. 7
    The recommended manual action

    The smallest useful next step a human should take — a draft you edit, not a sequence that fires. Often the right first move isn’t a pitch at all.

  8. 8
    Approval state

    Nothing proceeds until you decide. Approve, hold, or discard — your decision and reasoning are logged with the pack, permanently.

§ 04The lineup

Four familiar tools. Four missing pieces.

Each of these does something. None of them can hand you an opportunity you could defend to a skeptical colleague — quote, source, reasoning, and all.

File A — Lead databases

A million rows, zero reasons.

Contact data ages by the week, and even when a row is accurate, nothing in it tells you why this company — or why today. The thinking is left entirely to you.

No why
File B — Scrapers

Volume without provenance.

Bulk collection strips context on the way in. You inherit the data risk, the staleness, and the judgment calls — with no record of where anything came from.

No provenance
File C — Cold-email automation

Sends first, thinks never.

Sequences optimize for throughput, so every recipient becomes a variable in someone else’s funnel. Your domain reputation is the fuel it burns.

No brakes
File D — Generic AI agents

Confident, and unaccountable.

Ask for prospects and you’ll get fluent, plausible answers — with no citations, no capture times, and no way to tell the found from the fabricated.

No evidence

Vox Prospect finds opportunities you can defend with proof.

Every pack arrives with the quote, the live source, the reasoning, the risk — and a human approval gate in front of any contact. If a case can’t be defended with evidence, it never reaches your queue.

Proof on file
§ 05The approval gate

Nothing leaves without your signature.

The gate isn’t a compliance checkbox bolted on at the end. It’s the architecture: the engine can find, argue, and draft — it cannot contact. Only you can open the gate, one pack at a time.

  1. V.1
    No blind sending.

    Outreach drafts stay drafts until you’ve read them, edited them, and approved them. There is no “send all” button — deliberately.

  2. V.2
    No autonomous contact. Ever.

    Vox Prospect never messages a prospect on its own — not on a schedule, not on a trigger, not “just this once.” The capability does not exist in the product.

  3. V.3
    Approved source lanes only.

    Hunters read the public lanes you’ve enabled and nothing else. You can see the lane list, edit it, and shut any lane off mid-run.

  4. V.4
    Evidence preserved.

    Every quote is kept with its URL, capture time, and an archived copy — so a claim you approved in July can still be verified in December.

  5. V.5
    A full decision trail.

    Each pack records what was found, when, what was recommended, and what you decided. If a client or colleague asks “why them?” — you can show the file.

the operator
Approved by — the only signature that opens the gate
Approval queue · Run VP-0158 3 awaiting your decision
Copperline Freight Co-opVP-0158-B

Two-line note answering the contract objection; pilot terms up front.

Awaiting approval
Harrowgate BinderyVP-0158-E

Reply in-thread with the fulfillment workaround they asked about. No pitch.

Awaiting approval
Saltmarsh ProvisionsVP-0158-F

Call brief for Thursday’s trade meet — they’ll be at booth 40.

Awaiting approval
Messages sent by Vox Prospect without approval, all-time: 0
§ 06Who runs it

Built for people who actually need customers.

Not marketing theater. Four operators, four runs, four queues of proof.

Case 01 — SaaS founder

Finding teams mid-complaint about the incumbent.

She pastes her own site. The intent map keys on frustration with the tool she replaces; hunters surface threads where teams are describing the exact failure — in their own words, this month.

Surfaced“Renewal is in Sept and nobody on the team will fight for it…”
SRC forum.example/t/9917 · fit 89 · awaiting approval
Case 02 — Local service business

Spotting companies with visible response gaps.

A commercial cleaning outfit runs its own site. Packs come back on nearby businesses with unanswered reviews, dead contact forms, and listings gone stale — documented gaps, not guesses.

Surfaced4 unanswered 2-star reviews in 60 days; form returns an error.
SRC listings.example/marrowgate-inn · fit 78 · awaiting approval
Case 03 — Agency

Shipping clients proof, not CSVs.

The agency runs each client’s site and delivers opportunity packs with the evidence attached. Every recommendation survives the client’s scrutiny, because the client can click the source.

SurfacedClient deck: 7 packs, 7 live citations, 0 rows of filler.
Run VP-0171 · delivered with decision trail
Case 04 — Portfolio operator

Testing demand across five products at once.

Same engine, five URLs. He reads where real, cited demand shows up first — and puts the quarter behind the product whose queue filled with the strongest proof, before committing a roadmap.

SurfacedProduct C: 11 packs, median intent 71. Products A/B/D/E: under 4.
Runs VP-0163–0167 · decision made on evidence
Exhibit A

§ 07 — Closing argument

Give Vox Prospect a product.
Get back proof-backed opportunities.

One URL to start. A short queue of argued, cited, human-approved cases to finish. The rest of the internet can keep the spreadsheet.

Start with a website → See opportunity pack

No contact lists to import · no blast sends · nothing contacted without approval