Build an ICP model from your closed-won customers in ChatGPT

Built by: Lusha
Time to build: 1 min
Difficulty: Easy
Tools: ChatGPT ▪ Lusha
Type: Prompt

Example outputs in this play are illustrative — they reflect the structure, fields, and format of real Lusha connector output, but were not pulled from a live session. Run the prompt with your own closed-won customer list to see live results.

Most ICPs are built from assumptions.

A sales team says “we sell to B2B SaaS.” Marketing says “mid-market RevOps.” RevOps adds a few filters. Then everyone builds campaigns around a profile that may or may not match the customers who actually convert, retain, and expand.

Your closed-won customers are a better starting point.

This prompt uses Lusha in ChatGPT to analyze your best customer accounts, enrich each company, identify shared firmographic and signal patterns, find lookalike companies, and turn the results into an actionable ICP model. Instead of guessing what good-fit looks like, you build the model from accounts that already worked.

How to start

Open in Claude Open in ChatGPT
1

Open Lusha in ChatGPT

Go to Lusha in ChatGPT and click “Start chat.” Every conversation started this way is automatically Lusha-enabled.

2

Or invoke Lusha in any existing conversation

Type @Lusha in the prompt bar and select Lusha from the dropdown. Unlike Claude, Lusha does not activate automatically in every ChatGPT conversation. You must invoke it every time.

3

Add your closed-won customer list

Copy the prompt below, add at least five closed-won customer domains or LinkedIn company URLs, and send. Lusha enriches the accounts and finds patterns ChatGPT can turn into an ICP model.

The prompt

Start from Lusha in ChatGPT or type @Lusha before sending.

@Lusha Build an ICP model from my closed-won customers.

CLOSED-WON CUSTOMER SEEDS:
Add 5-100 customer company domains or LinkedIn company URLs.

1. [customer domain or LinkedIn company URL]
2. [customer domain or LinkedIn company URL]
3. [customer domain or LinkedIn company URL]
4. [customer domain or LinkedIn company URL]
5. [customer domain or LinkedIn company URL]

CUSTOMER CONTEXT:
Why these customers are strong examples:
[high ACV / fast sales cycle / high retention / expansion /
strategic segment / successful use case / other]

MY PRODUCT:
[One sentence describing what you sell and the problem
it solves]

TARGET BUYERS:
Primary personas: [titles or personas]
Relevant departments: [Sales / Marketing / RevOps / IT /
Operations / Customer Success / Finance / HR / other]

EXCLUSIONS:
Do not include:
- Existing customers: [domains, if any]
- Competitors: [domains, if any]
- Poor-fit industries: [industries, if any]
- Poor-fit company sizes: [sizes, if any]
- Poor-fit regions: [regions, if any]

Using Lusha, do the following:

1. VALIDATE THE SEED CUSTOMERS
   Match each seed company to a verified Lusha company
   profile.

   Return:
   - Company name
   - Domain
   - Industry
   - Employee count
   - HQ location
   - Revenue range if available
   - Company LinkedIn if available
   - Match status

   If fewer than 5 seed companies can be verified, ask me
   for more seeds before continuing.

2. FIND THE SHARED ICP PATTERNS
   Analyze the verified seed companies and identify
   recurring patterns across:
   - Industry
   - Company size
   - Region
   - Revenue range, if available
   - Business model, if available
   - Tech stack, if available
   - Growth stage or funding context, if available
   - Relevant departments or functions

   Separate strong patterns from weak or inconsistent ones.

3. CHECK COMMON BUYING SIGNALS
   For the seed companies, check recent and historical
   signals if available.

   Look for patterns across:
   - Hiring surges
   - Hiring surges by relevant department
   - Funding events
   - Leadership changes
   - Headcount growth or reduction
   - IT spend changes
   - Website traffic changes
   - Commercial activity news
   - Corporate strategy news
   - Product activity news
   - Risk news
   - Intent topics related to my product, if available

4. BUILD THE ICP MODEL
   Create an explainable ICP model based only on the Lusha
   data returned and the context I provided.

   Return:
   - Core ICP definition
   - Best-fit company profile
   - Best-fit buyer profile
   - Strong fit indicators
   - Weak fit indicators
   - Disqualifiers
   - Signals that suggest good timing
   - Signals that suggest lower urgency

   Do not present the model as a guarantee of conversion.

5. FIND LOOKALIKE ACCOUNTS
   Use the verified seed customers to find lookalike
   companies that resemble the ICP model.

   Do not include existing customers, competitors, or
   excluded segments.

   For each lookalike account, return:
   - Company name
   - Domain
   - Industry
   - Employee count
   - HQ location
   - Revenue range if available
   - Relevant signal, if available
   - Why it matches the ICP model

6. SCORE EACH LOOKALIKE ACCOUNT
   Create an explainable fit score from 1-100 using only
   the Lusha data returned.

   Break the score into:
   - ICP match: 40 points
   - Timing signals: 25 points
   - Persona relevance: 20 points
   - Actionability: 15 points

   If there is not enough data to support the score, mark
   the confidence as low and explain what is missing.

7. FIND STARTING CONTACTS
   For the top 5 lookalike accounts, find 1-2 relevant
   contacts matching the target buyer personas or departments.

   Return:
   - Name
   - Current title
   - Department
   - Seniority
   - Location
   - LinkedIn profile if available
   - Verified business email availability
   - Direct or mobile phone availability
   - DNC status if available

8. CREATE THE GO-TO-MARKET RECOMMENDATION
   Recommend how to use this ICP model.

   Return:
   - Best campaign angle
   - Best outbound trigger
   - Best personas to prioritize
   - Best account segment to start with
   - Messaging angle
   - One example opening line under 30 words

   Do not invent signals, tools, vendors, internal projects,
   buying intent, contacts, emails, phone numbers, or scores.

9. OUTPUT FORMAT
   Return:
   - Seed customer validation
   - Shared ICP patterns
   - Common signals
   - ICP model
   - Lookalike account table
   - Fit score and confidence
   - Starting contacts, if available
   - GTM recommendation
   - Excluded accounts and why

If Lusha cannot verify a company, contact, or signal, mark it
clearly rather than guessing.

What you’ll get back

 

An ICP model built from your closed-won customers, plus lookalike accounts, fit scores, signals, and starting contacts. Here’s what the output looks like:

Closed-won ICP model — Lusha

FieldValue
Seed customers12 submitted · 10 verified · 2 need review
Core ICP patternB2B SaaS · 200–1,000 employees · North America · RevOps-led teams
Common signalHiring growth in Sales and RevOps · appears across strongest customer matches
Top lookalike[Company A] · fit score 88/100 · high confidence
Starting contactR.M. · VP Revenue Operations · verified email available · mobile available
GTM recommendationStart with RevOps-led SaaS accounts showing sales hiring growth

Example outputs in this play are illustrative — they reflect the structure, fields, and format of real Lusha connector output, but were not pulled from a live session. Run the prompt with your own closed-won customers to see live results.

 

Why use Lusha in ChatGPT to build an ICP model

 

A strong ICP should be built from evidence, not assumptions.

Closed-won customers show which accounts already understood the value, had the right pain, and were willing to act. Lusha helps turn that customer base into a repeatable targeting model by enriching the accounts, finding shared patterns, checking signals, and surfacing similar companies.

ChatGPT adds the interpretation layer. It turns the Lusha data into an explainable ICP model, fit scores, recommended account segments, and practical campaign angles. That gives sales and marketing a shared view of what good-fit looks like and where to focus next.

The result is an ICP model that is easier to explain, easier to test, and easier to activate.

Lusha data is sourced and used in accordance with Lusha’s Privacy Policy and Terms of Use. Lusha is GDPR compliant and covers contacts across North America, EMEA, and APAC.

FAQ

  • How many closed-won customers do I need?

    Use at least five seed companies. More seeds usually give ChatGPT more signal to compare, but the prompt asks Lusha to stop and ask for more if fewer than five companies can be verified.

  • Is this a predictive model?

    No. ChatGPT creates an explainable ICP model from the Lusha data returned and the context you provide. It helps prioritize, but it does not guarantee conversion.

  • Can I use this for campaign planning?

    Yes. The output includes ICP patterns, lookalike accounts, target personas, signal themes, and campaign angles that sales and marketing can use together.

  • What if my closed-won customers are from different segments?

    The prompt separates strong patterns from weak or inconsistent ones. If the seeds represent multiple segments, it should call that out rather than forcing one generic ICP.

  • Can I exclude existing customers and competitors?

    Yes. Add exclusions in the prompt so the lookalike account list does not include customers, competitors, open opportunities, or poor-fit segments.

Related Prompts

Ready to run this?

One data connection. Works in Claude, ChatGPT, your CRM, or any agent you build.

Get free access Read the Docs