Skip to content
Hands-on AI Playbook

Workflow Example: Guided Prospect Research

AI involvement: Makes bounded decisions with your method — judges each prospect against the persona’s criteria and works out how to navigate LinkedIn — inside a fixed structure and report template.

What This Workflow Type Is

Section titled “What This Workflow Type Is”

A Guided workflow is one where you give the AI bounded decisions, with your method. You set the structure and the methodology; the AI uses it to make decisions on your behalf, and those decisions decide what happens next. Here the method is a buyer persona: its criteria tell the AI how to judge each prospect. The AI’s judgment decides who advances to the report and who is dropped, and the AI also works out how to navigate LinkedIn to find candidates worth judging. The steps and the report template stay fixed; the decisions are bounded by your rules, not open-ended. That makes it Guided — not Deterministic (the AI’s output changes what happens next) and not Autonomous (you can list the steps, not just the goal). It is Augmented because of its human checkpoint: after analyzing the persona, it pauses for you to confirm the targeting criteria before it searches, and you review the report at the end.

Characteristics

Section titled “Characteristics”
  • Repeatable — the same structure every time, with any buyer persona
  • Predictable where it matters — output format and evaluation criteria are defined in advance; the AI’s decisions are bounded by them
  • Delegatable — anyone with the persona file can run it and get consistent results
  • Automatable — can run on a schedule or be triggered by a pipeline

When to Use

Section titled “When to Use”

Use this pattern when you have a task that:

  • Follows clear, documented rules or criteria
  • Takes structured input and produces structured output
  • Needs the AI to find its own way through a system you can’t script step by step (a website, a search interface)
  • Repeats on a regular cadence (weekly prospecting, monthly scans)

Example Scenario

Section titled “Example Scenario”

The problem: A sales leader needs to identify LinkedIn prospects that match a specific buyer persona. The research process is always the same — search by title and industry, evaluate against persona criteria, document findings in a consistent format — but manually doing it takes 45-60 minutes per batch. The criteria don’t change between runs; only the prospects found are different.

The solution: A workflow prompt that takes a buyer persona as input, runs a set LinkedIn research sequence, and produces a structured prospect report with engagement recommendations. The evaluation criteria come directly from the persona file, and the AI uses them to decide which prospects make the report. It also decides how to navigate LinkedIn to find candidates worth evaluating.

Building Blocks

Section titled “Building Blocks”
Building BlockTypeDescriptionSource
linkedin-prospect-researchPromptWorkflow that finds and qualifies 5 LinkedIn prospects against a buyer personaView on GitHub
buyer-persona-revenue-leader-rachelPromptExample buyer persona used as input to the research workflowView on GitHub

How It Works

Section titled “How It Works”
graph LR
    A[Buyer persona<br>defines criteria] --> B[Workflow prompt<br>searches LinkedIn]
    B --> C[Evaluate prospects<br>against persona criteria]
    C --> D[Structured report<br>with 5 qualified prospects]

Step-by-step:

  1. Provide the buyer persona — the persona file defines the target titles, industries, company sizes, pain points, and trigger events. These become the evaluation criteria.
  2. Workflow analyzes the persona — extracts job titles, industry, company size, seniority, location, and exclusion criteria.
  3. Workflow searches LinkedIn — uses the persona criteria as search filters, systematically reviewing results.
  4. Workflow evaluates each prospect — checks role match, company fit, engagement signals, and accessibility (mutual connections, shared groups). Selects the top 5.
  5. Workflow generates the report — documents each prospect with profile URL, title, company, persona match reasons, and a specific engagement hook. Includes a summary with selection rationale, common themes, and priority ranking.
  6. Quality check — verifies all 5 prospects match criteria, URLs are complete, engagement recommendations are specific (not generic), and the output is properly formatted.

The Workflow in Detail

Section titled “The Workflow in Detail”

The prompt encodes a five-step sequence that the AI executes in order:

StepWhat HappensWhat Decides the Next Step
1. Analyze personaExtract titles, industry, size, location, pain points, triggers, then pause for you to confirm the targeting criteriaYou — criteria are predefined in the input file, and nothing is searched until you confirm them
2. Access LinkedInNavigate and confirm authenticationYou — binary check, logged in or stop
3. Search and evaluateApply persona criteria as search filters, score prospectsThe AI’s judgment, by your rules — it scores each prospect against the persona (title match → company fit → engagement signals → accessibility) and that score decides who advances to the report. It also works out which searches to run, which profiles to open, and when to page further
4. Document prospectsCapture name, URL, title, company, match reasons, engagement hookYou — output fields are prescribed
5. Generate reportFormat into template with summary, themes, and priority orderYou — template is rigid, same structure every run

Step 3 is the only step where an AI decision sets what happens next, and one is enough: a workflow’s autonomy level is the highest level any of its steps reaches.

The prompt also includes explicit error handling for three failure modes (fewer than 5 matches, LinkedIn access restricted, incomplete persona), so the workflow handles exceptions without human intervention.

The Buyer Persona

Section titled “The Buyer Persona”

The included example persona — “Revenue Leader Rachel” — demonstrates the kind of structured input this workflow expects. Key fields the workflow extracts:

  • Target titles: SVP/VP of Revenue, CRO, VP of Customer Solutions, VP/Head of Operations, Chief Customer Officer
  • Company context: B2B SaaS, 200-2,000 employees, Series B through public
  • Location: Major metro areas (NYC, SF, Boston, Chicago)
  • Pain points: Scaling without headcount, cross-functional execution gaps, ROI on AI investments
  • Trigger events: Board asking about AI strategy, competitors announcing AI features, failed AI pilots

You can swap in any buyer persona that follows a similar structure. The workflow adapts to whatever criteria the persona defines.

Usage

Section titled “Usage”

This example uses two standalone prompts — the workflow and the buyer persona. Both are plain markdown files you can use with any AI tool that has web browsing.

  1. Open the linkedin-prospect-research prompt on GitHub
  2. Open the buyer persona on GitHub
  3. Copy the workflow prompt into Claude, ChatGPT, or Gemini (with web browsing enabled)
  4. Paste or attach the buyer persona as context
  5. The AI executes the workflow and produces the prospect report

Adapting This Example

Section titled “Adapting This Example”

LinkedIn prospect research is one application, but the pattern — the AI judging items by your criteria to decide which ones advance, inside a fixed structure — works for other tasks too:

  • Customer account scoring — score accounts against an ideal customer profile, produce a ranked list
  • Job candidate screening — evaluate resumes against a job description’s requirements, produce a shortlist with match reasons
  • Vendor evaluation — assess proposals against scoring criteria, produce a comparison matrix
  • Content audit — evaluate published content against brand guidelines, produce a compliance report
  • Competitive monitoring — check competitor websites against a tracking template, produce a change report

To adapt: identify the input criteria (the persona equivalent), the evaluation rules (how to score matches), and the output template (what the report looks like). When the AI’s judgment by those rules decides what happens to each item — advance, drop, send back — the workflow is Guided. If the AI only writes up results you’ve already selected, it’s Deterministic.

Section titled “Related”