How to Build a Simple AI Lead Scoring System with HubSpot, Google Sheets, and ChatGPT in 2026

If your small business gets a steady flow of website inquiries, form submissions, email replies, or booked calls, the hard part is no longer “getting leads.” The hard part is knowing which leads deserve attention first.

A simple AI lead scoring system for small business helps you rank new contacts based on fit, interest, and urgency so your sales team can follow up with the best opportunities first. You do not need a full enterprise CRM project to get started. With HubSpot, Google Sheets, and ChatGPT, you can build a practical first version that works well enough to test on real leads before investing in heavier automation.

TL;DR

• Use HubSpot to collect and store lead data.
• Use Google Sheets to clean, review, and test your scoring model.
• Use ChatGPT to assign a plain-English score, reason, and next action.
• Start with a simple 100-point model: 40 points for fit, 35 for intent, and 25 for urgency.
• Test the system on 25 recent leads before automating anything.

The Problem: Too Many Leads, Not Enough Sales Time

Small teams often waste hours chasing contacts who are not ready to buy. Some people are doing early research. Some are outside your service area. Some do not have the budget. Others may be a great fit but need a faster response than your team can provide if every inquiry is treated the same.

Lead scoring solves this by sorting leads into priority groups. Think of it like sorting mail into three piles: urgent, later, and ignore for now. You still keep the information, but you do not spend the same amount of time on every item.

The business goal is simple: rank leads so sales follows up with the best opportunities first. For a small business, even saving 3 to 5 hours per week can matter. That time can be redirected toward booked calls, proposals, customer follow-up, or higher-quality outreach.

Who This Setup Is For

This setup is a good fit for solo operators, consultants, agencies, home service businesses, and 5-50 person teams that want better lead prioritization without buying an enterprise platform.

It works especially well if leads already come through one or more of these sources:

• HubSpot forms
• Website contact forms
• Calendly or other booking tools
• Email campaigns
• Manual sales imports
• Paid ad landing pages

This is also a reasonable starter option for teams that are not ready for enterprise tools like Salesforce Einstein or dedicated predictive scoring platforms. HubSpot does offer manual and AI-assisted scoring features, and those may be the better long-term choice once your company has enough historical conversion data. But for many small teams, a spreadsheet-based pilot is the fastest way to learn what actually predicts a good lead.

You should be comfortable creating or editing HubSpot properties, exporting or syncing contacts to Google Sheets, and copying prompts into ChatGPT. This article is practical business technology guidance, not legal, financial, or certified IT advice.

The Simple Tool Stack and What Each Tool Does

The goal is not to create a complex AI system. The goal is to create a repeatable workflow that makes lead follow-up more consistent.

HubSpot acts as your source of truth. Google Sheets acts as the testing and review layer. ChatGPT acts as the scoring assistant. Zapier, Make, or n8n can later connect the pieces so the process runs automatically.

Step 1: Define What Makes a Good Lead Before Using AI

Do not start by asking ChatGPT to “score my leads.” Start by defining what a good lead means for your business.

For most small businesses, a useful scoring model includes 5 to 7 factors:

• Budget fit: Can the lead afford your typical project, service, or package?
• Company size: Is the organization the right size for your offer?
• Service need: Are they asking for something you actually provide?
• Urgency: Do they need help now, this quarter, or “someday”?
• Location: Are they inside your service area or target market?
• Engagement: Did they visit high-intent pages, reply to emails, or request a consultation?
• Decision-maker status: Are they the owner, executive, manager, or authorized buyer?

A simple 100-point model is enough for a first version:

• Fit: 40 points for budget, company size, location, and service match.
• Intent: 35 points for demo requests, pricing page visits, form answers, replies, and content engagement.
• Urgency: 25 points for timeline, stated pain, deadline, or current buying stage.

For example, a lead that visits your pricing page, requests a demo, and says they need help within 30 days should score higher than someone who only signs up for a newsletter. A homeowner requesting an estimate inside your service area should score higher than a student asking for research help. A business owner who says “we have budget approved and need a vendor this month” should score higher than a vague inquiry with no timeline.

You should also define negative signals:

• Student or academic research inquiry
• Competitor inquiry
• Outside service area
• Fake or suspicious email address
• No business website when one would normally be expected
• Service request that does not match what your company sells
• Budget far below your minimum engagement

Before automation, sales and marketing should agree on what counts as a Sales Qualified Lead. If one person thinks a “hot lead” means any form fill and another thinks it means budget plus timeline plus decision-maker status, your scoring system will create confusion.

Step 2: Prepare HubSpot and Google Sheets Data

AI scoring is only as useful as the data you provide. If HubSpot has incomplete, duplicated, or vague contact records, ChatGPT will make weaker recommendations.

Create or Confirm HubSpot Properties

In HubSpot, create or confirm fields such as:

• Lead Source
• Company Size
• Service Interest
• Budget Range
• Timeline
• Last Engagement
• Lead Score
• AI Score Reason

You may already have some of these fields under different names. Use the fields that match your current process, but try to make the data structured. Dropdowns are usually better than open text when you want consistent reporting.

Set Up Your Google Sheet

Export recent contacts from HubSpot or sync new form submissions into Google Sheets. Your sheet should include columns such as:

• Name
• Company
• Email
• Lead Source
• Form Answers
• Website Activity
• Lifecycle Stage
• Deal Outcome
• Sales Notes
• AI Lead Score
• AI Score Reason
• Recommended Follow-Up

Clean the sheet before scoring. Remove duplicates. Standardize industries. Fix blank service-interest fields where possible. Label won and lost deals if you have that history. Even a small amount of clean historical data can help you compare whether the AI score matches real outcomes.

The limitation is straightforward: weak CRM data produces unreliable AI scores. ChatGPT can organize and interpret information, but it cannot magically know whether a lead is qualified if the source data is missing or misleading.

Step 3: Use ChatGPT to Score Leads Consistently

Once your data is clean enough for a pilot, use a consistent prompt. The prompt should tell ChatGPT what scoring factors matter, how points are assigned, and what format to return.

Reusable ChatGPT Lead Scoring Prompt

You are helping score sales leads for a small business.

Use this 100-point model:
- Fit score: 0-40 points based on budget fit, company size, location, service match, and decision-maker status.
- Intent score: 0-35 points based on form answers, pricing page visits, demo requests, email engagement, and website activity.
- Urgency score: 0-25 points based on stated timeline, pain level, deadline, and readiness to speak with sales.

Score bands:
- 80-100: Hot lead
- 60-79: Nurture
- 40-59: Low priority
- Under 40: Do not pursue yet

Negative signals include student research, competitor, outside service area, fake email, no business website when one is expected, poor service fit, or budget below our minimum.

Return the result as JSON with these fields:
{
  "total_score": 0,
  "fit_score": 0,
  "intent_score": 0,
  "urgency_score": 0,
  "grade": "",
  "score_reason": "",
  "recommended_follow_up": ""
}

Keep the score_reason to one sentence so a human can audit it.

Lead data:
Name: [Name]
Company: [Company]
Email: [Email]
Lead Source: [Lead Source]
Company Size: [Company Size]
Service Interest: [Service Interest]
Budget Range: [Budget Range]
Timeline: [Timeline]
Website Activity: [Website Activity]
Form Answers: [Form Answers]
Lifecycle Stage: [Lifecycle Stage]
Sales Notes: [Sales Notes]

You can paste one lead at a time for a manual test. For a more structured workflow, paste rows from Google Sheets and ask ChatGPT to return a table with the same fields.

Example Score

Here is a representative example:

{
  "total_score": 86,
  "fit_score": 35,
  "intent_score": 31,
  "urgency_score": 20,
  "grade": "Hot lead",
  "score_reason": "The lead matches the target service, has budget range alignment, visited the pricing page, and requested help within 30 days.",
  "recommended_follow_up": "Assign a sales task today and send a meeting link with two suggested time windows."
}

The one-sentence reason matters. Your team should be able to audit the recommendation quickly. If the reason does not make sense, the score should not be trusted.

Also be careful with sensitive personal data. Before sending customer or prospect information into any AI tool, your business should review its privacy, security, and contractual requirements. In many cases, you can test the workflow with limited business data, anonymized examples, or fields that do not include sensitive details.

Step 4: Put the Score Back Into HubSpot and Trigger Follow-Up

Start manually. Import scored rows from Google Sheets back into HubSpot as a first test. Map the AI score to your Lead Score property and the explanation to your AI Score Reason property.

Once you trust the scoring model, you can automate the workflow with Zapier, Make, or n8n:

1. A new HubSpot contact is created from a form, meeting booking, or manual entry.
2. The automation sends selected lead fields to ChatGPT.
3. ChatGPT returns a score, grade, reason, and recommended follow-up.
4. The automation writes the result to Google Sheets for review history.
5. The automation updates the HubSpot contact property.
6. HubSpot uses the score to trigger sales tasks, nurture emails, or list membership.

Create HubSpot Views

Once scores are back in HubSpot, create simple saved views:

• Hot Leads Over 80: Sales should review these first.
• Needs Nurture 60-79: These leads may need education, email follow-up, or retargeting.
• Low Fit Under 40: These leads should usually be suppressed from direct sales follow-up.

Add Simple Workflows

Then build light workflows around the score bands:

• If AI Score is over 80, assign a sales task.
• If AI Score is 60-79, enroll the lead in a nurture sequence.
• If AI Score is under 40, exclude the lead from active sales outreach unless a human overrides it.

A practical example: if a lead has an AI Score above 80, visited the pricing page, but has not booked a meeting, HubSpot can trigger a task for the sales rep and send a meeting-link email. That is a simple, useful workflow because it responds to both fit and behavior.

Limitations, Costs, and When This Will Not Work

This setup is useful, but it is not magic. ChatGPT scoring can help your team prioritize, but it should not replace human judgment for large deals, unusual opportunities, regulated industries, or sensitive decisions.

HubSpot native AI or predictive scoring may be a better option if your business has enough historical conversion data and the right HubSpot plan. Native scoring can use CRM and engagement data directly inside the platform, which may reduce manual exports and integration complexity.

For a starter setup, expect costs from free to roughly $20-$100 per month depending on your ChatGPT plan, HubSpot tier, and automation volume. Costs can rise if you process many leads, use premium automation tools, or need custom integrations.

This approach may not work well if:

• Your lead data is mostly blank or inconsistent.
• Your team has not defined what a qualified lead means.
• You sell many unrelated services with different qualification rules.
• Your sales team ignores CRM updates.
• You automate before testing the scores manually.

Review the results every 30 days. Compare high scores against booked calls, closed deals, and sales feedback. If low-scoring leads keep turning into good customers, adjust the model. If high-scoring leads do not respond or do not close, inspect which factors are being overweighted.

What to Do Now

The best next step is not full automation. Build one Google Sheet scoring template and test it on 25 recent leads.

1. Export 25 recent leads from HubSpot.
2. Add columns for fit, intent, urgency, total score, reason, and next action.
3. Use the ChatGPT prompt above to score each lead.
4. Compare the scores against what actually happened: booked call, no response, won deal, lost deal, or bad fit.
5. Adjust your scoring rules with sales feedback.
6. Only automate after the scores make sense.

A simple AI lead scoring system for small business should make sales follow-up clearer, not more complicated. Start with clean data, a practical scoring model, and human review. Once the pattern works, connect HubSpot, Google Sheets, and ChatGPT with automation so your team can spend less time sorting leads and more time talking to the right prospects.