Last month I was reviewing a consulting client's calendar. Out of 18 discovery calls booked that month, 11 of them went nowhere. Wrong budget. No real urgency. Decision-maker not on the call. One guy just wanted free advice.

That is 11 hours of prep, context-switching, and follow-up for zero revenue. At even a conservative billing rate of $150/hour, that is $1,650 in lost earning potential in a single month. Multiply that across a year and you are looking at close to $20,000 wasted on calls that should never have happened.

This is not a time management problem. It is a qualification problem. And the fix is not a better calendar, it is a smarter gate.

Here is how I build AI lead qualification bots for consulting clients, and how you can do the same.

The Real Cost of Unqualified Discovery Calls

According to a 2025 Entrepreneur analysis, if 25% of your leads are clearly unqualified, that is 25% of your time budget thrown away before any real work begins. For solo consultants running on 20 billable hours per week, that can mean 5 hours every single week gone to conversations that produce nothing.

Research from Landbase puts it plainly: only 27% of leads sent to sales pipelines are actually qualified. That means roughly 73% of people requesting your time are not your ideal client yet.

The average consultant I work with is taking 6 to 10 discovery calls per month. If even half of those are poorly qualified, you are looking at 3 to 5 hours of dead-end conversations, plus the mental load of each one. That adds up fast.

The solution is not to stop taking calls. It is to stop taking the wrong calls.

What a Lead Qualification Bot Actually Is

Let me be clear about what this is not. It is not a chatbot. It is not a live chat widget that answers FAQs. It is not an AI that replaces your first conversation.

A lead qualification bot is a structured scoring system. It intercepts prospects before they reach your calendar, asks them a focused set of questions, scores their responses against your ideal client profile using AI, and then routes them automatically based on their score.

High score: they see your Calendly link and book a call.

Mid score: they get a free resource and a follow-up in 30 days.

Low score: they receive a polite decline email pointing them toward a better fit.

This is not gatekeeping for its own sake. It is making sure your best energy goes to conversations that can actually become clients.

The 5-Question Framework I Use for Consultants

The questions vary slightly depending on the type of consulting, but the core framework stays consistent. You want to surface five things: budget alignment, timeline urgency, problem clarity, decision-making authority, and prior attempts.

Here is what that looks like in practice:

1. What is your approximate budget range for this project? Options: Under $1,000 / $1,000 to $3,000 / $3,000 to $10,000 / $10,000+

This single question eliminates a huge percentage of poor-fit leads immediately. If your minimum engagement is $5,000 and someone selects "Under $1,000," the system already knows.

2. What is your timeline for solving this problem? Options: I need help now (within 30 days) / Within 3 months / Within 6 months / Just exploring

Urgency is a proxy for readiness. Someone who is "just exploring" is not your next client. They might be a future one, but not now.

3. Describe the specific outcome you are trying to achieve. (Open text field, 2-3 sentences)

Vague answers score low. Specific, outcome-oriented answers score high. "I want to grow my business" is very different from "I need to reduce client churn from 40% to under 15% in the next quarter."

4. Who will be making the final decision on this engagement? Options: Me alone / Me and a partner / My manager/board needs to approve / Not sure

If the decision-maker is not in the room, the sales cycle gets complicated fast. This question surfaces that early.

5. Have you tried to solve this problem before? What happened? (Open text field)

Prior attempts reveal both seriousness and sophistication. Someone who has tried things and failed is often a better prospect than someone who has not thought about it at all. It also tells you what they have already ruled out.

You can add a sixth question around company size or industry if your ideal client profile is narrowly defined. But five focused questions is usually enough to score effectively.

How the AI Scoring Actually Works

Once someone submits the form, here is the workflow I build in n8n:

Step 1: Form submission triggers the workflow The form lives on Tally or Typeform. When submitted, it fires a webhook to n8n.

Step 2: n8n sends the responses to Claude or GPT-4o-mini with a scoring prompt The prompt includes:

A typical scoring prompt looks like this:

You are a lead qualification assistant for a consulting firm. 
Score the following prospect responses against this ideal client profile:
[ICP DESCRIPTION]

Score each answer from 1-2 based on fit:
1 = poor fit, 2 = strong fit

Return a JSON object with: total_score, score_breakdown, summary_reason

RESPONSES:
Budget: [response]
Timeline: [response]
Outcome: [response]
Decision maker: [response]
Prior attempts: [response]

GPT-4o-mini handles this well and keeps cost low. At roughly $0.15 per million input tokens, you could run thousands of qualifications per month before the cost becomes meaningful.

Step 3: n8n reads the score and routes accordingly

Step 4: All data writes to Supabase Every submission, score, score breakdown, and routing outcome goes into a Supabase table. This is the data layer that turns a simple bot into a business intelligence tool over time.

What to Do With the Scored Data

Most consultants set this up and then never look at the data. That is a missed opportunity.

The Supabase table I build includes columns for: submission date, budget range, timeline, problem description, decision-maker status, prior attempts, AI score, score breakdown, routing outcome, and eventual close status (manually updated).

Once you have 30 to 60 submissions, you can start running weekly reviews to identify patterns:

This data feedback loop is what separates a smart lead system from a glorified contact form. After 90 days you will have a clear picture of what your best clients look like before they get on a call with you.

The Tool Stack

Here is what this system runs on, all of it either free-tier or low-cost:

Tool Purpose Cost
Tally or Typeform Qualification form Free / $25/mo
n8n Workflow automation Self-hosted or $20/mo
Claude or GPT-4o-mini AI scoring Pay-per-use (very low)
Supabase Data storage Free tier covers most use cases
Calendly Booking (high-score route) Free / $10/mo

Total monthly cost for most solo consultants: $20 to $55, depending on whether you self-host n8n.

If you are currently wasting 4+ hours per month on unqualified calls, this system pays for itself the first week it runs.

Frequently Asked Questions

Will prospects be put off by having to fill out a form before booking a call?

The right prospects are not. In fact, many consultants find that a qualification form actually increases the perceived value of their time. If someone is serious, they will answer 5 questions. If filling out a form feels like too much effort, they were never going to become a client anyway.

What happens to leads who score 4 to 6? Are they lost?

No. The nurture sequence keeps them warm. Sending a genuinely useful free resource with a 30-day follow-up often converts a mid-score lead once their timeline or budget situation changes. Some of my clients' best long-term clients started as score 5 leads.

Do I need technical skills to build this?

You need to be comfortable with n8n's workflow builder (drag and drop, no code) and Tally or Typeform. The Supabase table setup takes about 15 minutes with a basic guide. If you want someone to build the whole thing for you, that is exactly what I do.

How long does it take to set up?

For a consultant with a clear ideal client profile, I can have this running in 3 to 5 hours. Most of the time is in defining the ICP and writing the scoring prompt, not the technical build.

What AI model should I use for scoring?

GPT-4o-mini is my default. It is fast, cheap, and more than capable for structured scoring tasks. Claude Haiku is a solid alternative. You do not need GPT-4o or Claude Sonnet for this. Save those for heavier reasoning tasks.

Sources

Want to see the full picture of how lead qualification fits into a broader consulting automation stack? Read the pillar post: AI Automation for Consultants: What Actually Works in 2026.

Ready to stop burning hours on discovery calls that go nowhere? Let us build your lead qualification system. Get in touch here.

Need AI automation for your business?

I build custom AI systems that save 20+ hours/week. Book a free discovery call to see what's possible.

Book a Discovery Call