How to Build a $0 Lead Scoring System With Activepieces + Free LLMs (Complete Setup Guide)

Introduction: From Reading About AI Scoring to Actually Having It Running

You've read about AI lead scoring. Maybe you've even read articles explaining why you need it.

This isn't another explainer.

This is the step-by-step build guide showing you:

• Which free accounts to create (with direct links)

• The exact workflow configuration in Activepieces

• The LLM prompt that scores leads accurately

• How to connect it all to HubSpot without code

Time investment: 3-4 hours this weekend
Technical skill needed: If you can copy/paste and follow screenshots, you're ready
Cost: $0 until you process 1,000+ leads/month

The Free Tool Stack (Why These Specific Tools)

Key decision: Activepieces vs Make?

• Choose Activepieces if you want built-in AI steps (easier)

• Choose Make if you're already using it for other workflows

This guide uses Activepieces as primary, with Make alternatives noted.

Understanding the System You're Building (5-Minute Overview)

Input: Lead submits form → Name, Email, Company, Message
Process:

1. Webhook receives data

2. Optional: Enrich with company info

3. LLM analyzes fit + intent

4. Returns JSON: {score: 8, route: "book_call", confidence: "high"}

5. System routes based on score

Output:

• Score 8-10 → Email with calendar link

• Score 5-7 → Add to nurture in HubSpot

• Score 0-4 → Tag as "low-fit" for later review

Step 1 — Set Up Your Free Accounts (15 Minutes)

Account Creation Checklist:

HubSpot CRM (Free Forever Plan)

1. Go to hubspot.com/products/crm

2. Click "Get started free"

3. Complete signup

4. Navigate to Settings → Properties → Create custom property: ai_lead_score (number field)

Activepieces

1. Visit activepieces.com

2. Sign up for free account

3. Verify email

4. You get 1,000 tasks/month free tier

OpenRouter (Free LLM Access)

1. Go to openrouter.ai

2. Create account

3. Navigate to Keys → Generate API key

4. Copy key (you'll paste this into Activepieces)

5. Check "Free Models" tab - note z-ai/glm-4.5-air:free

Step 2 — Create Your Lead Capture Form in HubSpot

HubSpot forms feed directly into your automation:

1. HubSpot → Marketing → Forms → Create form

2. Add fields:

   • First Name (required)

   • Last Name (required)

   • Email (required)

   • Company Name (required)

   • Job Title (required)

   • Message/How can we help? (long text)

   • Company Size (dropdown: 1-10, 11-50, 51-200, 201+)

   • Timeline (dropdown: Urgent, 1-3 months, 3-6 months, Just researching)

3. Form Settings:

   • Thank you message: "Thanks! We'll review and respond within 24 hours."

   • Notification: Turn OFF (automation will handle routing)

4. Embed on site or use HubSpot's hosted page

5. Key setup: Go to Settings → Integrations → Private Apps → Create app → Give API access to contacts (read/write)

Step 3 — Build the Scoring Workflow in Activepieces

Creating Your First Flow:

In Activepieces Dashboard:

1. Click "Create Flow" → Name it "AI Lead Scorer"

2. Trigger: HubSpot - New Contact

   • Connect your HubSpot account

   • Select trigger: "Contact Created"

   • Test connection

   • Test by submitting data through the form

3. Step 2: HTTP Request to OpenRouter

   • Add step → HTTP Request

   • Method: POST

   • URL: https://openrouter.ai/api/v1/chat/completions

   • Headers:

     Authorization: Bearer YOUR_OPENROUTER_API_KEY
     Content-Type: application/json

• Body (copy this exactly, then customize):

json

     {
       "model": "z-ai/glm-4.5-air:free",
       "messages": [
         {
           "role": "system",
           "content": "You are a lead scoring AI. Analyze leads and return JSON with: score (0-10), route (book_call/nurture/archive), confidence (high/medium/low), reasoning (brief explanation). Score based on: company size fit, role seniority, timeline urgency, message intent quality."
         },
         {
           "role": "user",
           "content": "Score this lead:\nName: {{trigger.properties.firstname}} {{trigger.properties.lastname}}\nEmail: {{trigger.properties.email}}\nCompany: {{trigger.properties.company}}\nRole: {{trigger.properties.jobtitle}}\nCompany Size: {{trigger.properties.company_size}}\nTimeline: {{trigger.properties.timeline}}\nMessage: {{trigger.message}}"
         }
       ],
       "response_format": { "type": "json_object" }
     }

4. Step 3: Parse JSON Response

   • Add step → Code (it's JavaScript)

   • Code:

javascript

module.exports = {
  code: async (inputs) => {
    const response = JSON.parse(
      inputs.aiResponse.choices[0].message.content
    );

    return {
      score: response.score,
      route: response.route,
      confidence: response.confidence,
      reasoning: response.reasoning
    };
  }
};

Step 4: Router - Branch by Score

• Add Router step

• Create 3 branches:

  • Branch A: If score >= 8

  • Branch B: If score >= 5 AND score < 8

  • Branch C: If score < 5

7. Step 4 — Configure Routing Actions for Each Score Range

Branch A: High Score (8-10) → Book Call

Add HubSpot Action:

• Action: Update Contact

• Contact ID: {{trigger.id}}

• Properties to update:

  • ai_lead_score: {{parseJson.score}}

  • lifecyclestage: marketingqualifiedlead

  • lead_status: New

Add Gmail/Email Action:

• To: {{trigger.email}}

• Subject: Quick question about {{trigger.company}}

• Body: Hi {{trigger.firstname}},

  Thanks for reaching out! Based on what you shared, I think we could be a great fit.

  Would you be open to a quick 15-minute call this week? Grab a time here: [YOUR_CALENDLY_LINK]

  Best, [Your Name]

Branch B: Medium Score (5-7) → Nurture Sequence

Add HubSpot Action:

• Action: Add to Workflow

• Workflow: "Nurture Sequence" (create this in HubSpot as 3-5 email drip)

• Update properties:

  • ai_lead_score: {{parseJson.score}}

  • lead_status: Nurture

Branch C: Low Score (0-4) → Archive

Add HubSpot Action:

• Update properties:

  • ai_lead_score: {{parseJson.score}}

  • lead_status: Not a Fit

  • hs_lead_status: Unqualified

Optional: Send polite rejection

• "Thanks for your interest. Based on [reasoning], we might not be the best fit right now. Here are some resources that might help: [link to blog/guides]"

Step 5 — Test Your System With Real Data

Testing Checklist:

1. Submit a test lead through your HubSpot form:

   • Use your own email

   • Fill as "high-fit" lead: Founder role, 11-50 company, "Urgent" timeline, specific message

2. Watch Activepieces execution log:

   • Go to Runs tab

   • Click on the triggered flow

   • Verify each step executed

   • Check LLM response JSON

3. Verify HubSpot updates:

   • Open contact in HubSpot

   • Check ai_lead_score field populated

   • Verify lead status updated

   • Confirm you received routing email

4. Test edge cases:

   • Submit "low-fit" lead (just researching, junior role)

   • Submit "medium-fit" lead

   • Verify each routes correctly

The LLM Prompt That Actually Works (Copy This)

The prompt in Step 6 works because it:

• Specifies output format (JSON with exact fields)

• Defines scoring criteria (fit + intent signals)

• Requests reasoning (helps you debug/improve)

Prompt Customization Tips:

If you're B2B SaaS: Add to system prompt: "Prioritize leads from tech companies, CTOs/VPs, with budget mentions."

If you're agency: Add: "Score higher if message includes specific project details, realistic timeline, mentions referral."

If you're selling to enterprises: Add: "Require company size 200+, C-level or VP roles. Discount small businesses."

Logging Decisions for Continuous Improvement

Track Scoring Accuracy:

Create a Google Sheet tracker (optional but recommended):

1. Add Activepieces step after scoring:

   • Action: Google Sheets - Add Row

   • Spreadsheet: "Lead Scoring Log"

   • Columns: Date, Name, Email, Company, Score, Route, Confidence, Reasoning

2. Weekly review process:

   • Filter leads you manually followed up with

   • Compare AI score vs your assessment

   • Note patterns where AI was wrong

   • Refine prompt based on mismatches

Handling Edge Cases & Low-Confidence Scores

Fallback Logic:

Add this to your Router:

• If confidence == "low" → Route to manual review regardless of score

  • Create HubSpot task: "Review ambiguous lead - AI confidence low"

  • Assign to yourself

  • Include AI reasoning in task notes

Common low-confidence scenarios:

• Generic messages ("just checking it out")

• Missing key fields (no company size)

• Mixed signals (senior role but "just researching")

Solution: Build a 4th branch specifically for low-confidence → Human review queue

Free Tier Limits & How to Stay Under Them

Monthly Capacity on $0 Budget:

Cost Control Tactics:

1. Prevent duplicate scoring:

• Add filter in Activepieces: Only trigger if ai_lead_score is empty

• Code check:

javascript

  if (inputs.trigger.ai_lead_score) {
    return { skip: true };
  }

2. Batch low-priority leads:

• Score urgent leads immediately

• Batch-process "just researching" leads once daily

3. Cache enrichment data:

• If using company enrichment, store results in HubSpot custom field

• Don't re-enrich same domain

Weekend Implementation Checklist

Saturday Morning (2 hours):

• Create HubSpot, Activepieces, OpenRouter accounts

• Build HubSpot form with required fields

• Create custom property ai_lead_score in HubSpot

• Get OpenRouter API key

Saturday Afternoon (2 hours):

• Build Activepieces flow: Trigger → LLM call → Parse JSON → Router

• Configure 3 routing branches (book/nurture/archive)

• Connect HubSpot actions to each branch

• Set up email templates

Sunday (1 hour):

• Submit 5 test leads (varied fit levels)

• Verify each routes correctly

• Check HubSpot updates

• Confirm emails sent

Sunday Evening (30 min):

• Turn on live

• Monitor first real leads

• Document any issues

What to Do When You Outgrow Free Tiers

Signs you need to upgrade:

• Processing 200+ leads/month (Activepieces limit approaching)

• Need faster LLM responses (free models have rate limits)

• Want advanced enrichment (company funding, tech stack)

Upgrade path:

1. Month 1-3: Stay on free tiers, optimize prompt

2. Month 4-6: Add paid enrichment ($50/mo Clearbit/Apollo)

3. Month 7+: Upgrade Activepieces ($15/mo) or switch to paid LLM (GPT-4 for complex scoring)

When to hire vs automate more:

• If scoring 500+ qualified leads/month → Hire SDR, use system for pre-qualification only

• If scoring mostly unqualified → Keep automation, improve top-of-funnel

Next Steps: Make It Yours

This week:

• Build the system following steps 4-7

• Test with 10 real leads

• Adjust prompt based on results

Next week:

• Add enrichment (company data from Seamless.AI free tier)

• Create nurture sequences in HubSpot

• Track which routed leads convert

This month:

• Review 50 scored leads

• Calculate time saved vs manual qualification

• Refine ICP criteria in LLM prompt

You now have the exact blueprint. The only thing between you and an AI sales system is a weekend of setup.