AI Lead Analysis Engine: Evaluating 269 Businesses with a 21-Point Scoring System

The Problem

Effective outbound sales depends on reaching the right businesses with the right message at the right time. For that to work, leads need to be evaluated — not just collected.

The qualification process involves answering a set of questions about each potential target: Are they in the right industry? Do they have the right business model? What signals suggest they have the problem we solve? What's their digital maturity? Do they have the budget profile for our services?

Answering these questions manually for each lead is research-intensive work. A thorough evaluation takes 15–30 minutes per business. For a batch of 50 leads, that's a full day of work before a single outreach email is sent.

At that rate, outbound volume is constrained not by sales capacity but by research capacity. The bottleneck isn't the closer — it's the qualification process.

Our Approach

The core insight was that qualification follows consistent logic. The 21 criteria being evaluated don't change from lead to lead — only the inputs change. If the criteria and the reasoning are consistent, the evaluation process can be automated.

We designed a pipeline with three stages:

Data collection. For each business in the pipeline, structured data is assembled: company name, website, industry, size indicators, digital footprint signals (website quality, ad presence, review volume, social activity), and any available technographic data.

AI-driven evaluation. Each data record is passed through a structured prompt pipeline using the OpenAI API. The prompt encodes the 21 qualification criteria and instructs the model to evaluate each criterion, assign a score, and provide a brief rationale. The output is structured JSON — consistent format, programmatically parseable.

Prioritization and export. Total scores are calculated, leads are ranked, and the results are exported with individual criterion scores, overall fit ratings, and recommended outreach angles for the highest-priority targets. The output goes directly into Airtable for the sales team.

What We Built

Lead ingestion pipeline. A Python-based pipeline reads batch input files, normalizes business data into a consistent schema, and prepares records for evaluation.

21-point scoring engine. Each qualification criterion is defined in a structured prompt format with clear evaluation guidelines. The AI model evaluates each criterion independently, reducing the risk of one strong signal inflating the overall score.

Structured output parsing. Model outputs are parsed, validated, and stored as structured records. Malformed or incomplete responses trigger automatic retry logic before flagging for human review.

Airtable integration. Scored leads are pushed directly to Airtable via API, with views pre-configured for the sales team: all leads sorted by score, top-quartile leads filtered for immediate outreach, and leads flagged for review separated for human follow-up.

n8n orchestration. The end-to-end pipeline is orchestrated through n8n, enabling trigger-based execution (new batch file added → pipeline runs automatically) and monitoring with failure alerts.

Results

A batch of 269 businesses was evaluated, scored, and prioritized in a single pipeline run — work that would have taken 3–5 days manually. The scoring system produced differentiated outputs: high-fit leads with specific outreach angles, mid-tier leads for follow-up sequencing, and low-fit leads deprioritized without consuming sales attention.

The pipeline's value compounds over time. The same infrastructure processes every subsequent batch, and the scoring criteria can be refined as qualification logic improves — without rebuilding the system.

For any business running structured outbound, the principle applies: if your qualification criteria are consistent, the evaluation process can be automated. The variable that remains human is the relationship — and automating qualification frees your team to focus exactly there.