UIPrompt Library
Start Pro $14Account
Back to Income Lab
loopDraftReady now

Prospect research and proposal loop

Turns public business evidence into a small approval queue of tailored proposals and learns from actual replies.

Setup capital
Under $100
Monthly cost
$0–100
Setup time
3–6 hours
Weekly effort
5–10 hours
First dollar
1–4 weeks
Sales burden
Written outreach

User-selected scenario, not an expected outcome.

Every consequential external action pauses for explicit human approval.

Autonomy by business phase

Current level 3/5. Potential level 4/5 in 1–3 months.

1

Validate

AI-assisted

AI checks public fit signals against human-defined criteria.

2

Build

AI-assisted

AI prepares source notes and proposal drafts.

3

Acquire customers

AI-executed with approval

Every proposal remains queued until a human approves the send.

4

Deliver

AI-assisted

AI can prepare scope; a human owns commitments.

5

Collect and operate

Human-led

A human records outcomes and controls contracts and billing.

AI responsibilities

  • Research public fit signals
  • Cite sources
  • Draft tailored proposals
  • Summarize reply patterns

Human responsibilities

  • Approve every send
  • Honor opt-outs
  • Handle calls and negotiations
  • Approve scope and contract

Approval gates

Approval applies to the exact action once. It is not blanket permission.

Explicit approval

Outreach

Approve every recipient and exact message before sending.

Acquire customers
Explicit approval

Call

The human owns any discovery or closing call.

Acquire customers
Explicit approval

Contract

The human approves scope, terms, and signature.

Collect and operate
Runnable loop

Copy the complete asset

SETUP: Record the offer, ideal customer, disqualifiers, permitted sources, truthful proof, maximum batch size, opt-out policy, and CRM fields. Do not scrape private data or send messages automatically.

EACH CYCLE: (1) Research no more than ten prospects from public, relevant sources. (2) Save the source, date, fit signal, and reason to disqualify. (3) Score fit against the declared rubric; do not infer protected or sensitive traits. (4) Draft a short proposal that names one evidenced problem, one bounded deliverable, proof that actually exists, a price or next step supplied by the human, and a plain opt-out. (5) Queue each recipient and exact message for explicit approval. (6) The human sends approved messages and conducts all calls. (7) Log real replies, opt-outs, objections, meetings, and outcomes. (8) Revise the fit rubric or message only from observed data.

STOP when evidence is missing, opt-outs or bounces cross the chosen threshold, or truthful proof cannot support the proposal. ESCALATE regulated claims, sensitive data, procurement, security, and contract questions.

SETUP
1. Offer and ICP
2. Disqualifiers
3. Source and privacy policy
4. Proof boundaries
5. Batch and opt-out limits
6. CRM schema

RECURRING CYCLE
1. Research small batch
2. Cite and score fit
3. Draft proposals
4. Wait for approval
5. Human sends and calls
6. Log observed outcomes
7. Refine rubric

STATE TRACKING
1. Prospect
2. Source/date
3. Fit score
4. Disqualifier
5. Draft
6. Approval
7. Reply
8. Opt-out
9. Outcome

APPROVAL POINTS
1. Outreach — Explicit approval: Approve every recipient and exact message before sending.
2. Call — Explicit approval: The human owns any discovery or closing call.
3. Contract — Explicit approval: The human approves scope, terms, and signature.

STOP CONDITIONS
1. Missing fit evidence
2. Threshold breach
3. Invented proof would be required

ESCALATION RULES
1. Escalate privacy and regulated claims
2. Escalate procurement, security, and contract review

Evidence record

Projected scenarios never appear here as observed outcomes.

Evidence stateDraft

Loop is drafted; no prospect batch has been independently reproduced.

Experiment count0 recorded

No model test date recorded

Supported modelOpenAI GPT-5

Catalog baseline; exact version must be recorded at first test

Stop and escalate

  • Sources do not support fit
  • Opt-out or bounce threshold is exceeded
  • The proposal needs invented proof
  • Sensitive personal data
  • Regulated claims
  • Procurement, security, or legal review

Known failure modes

  • High-volume generic outreach
  • Fabricated personalization
  • Ignoring opt-outs
  • Overstating capabilities

Prerequisites and changelog

A consultant or small agency that needs a focused, evidence-based pipeline. · Research suitable prospects and prepare relevant proposals without automating unsolicited sends.

Before you start

  • Defined offer and ICP
  • Truthful proof boundaries
  • A lawful outreach channel
  • Prospecting
  • Offer positioning
  • Professional written communication

Changelog

  • 1.0.0 · 2026-07-12Initial launch workflow.