AI Lead Scoring API in India

AI Lead Scoring API in India

Sales teams often sit on hundreds of open leads with no reliable way to tell which ones are ready to buy. unoblox turns your CRM activity, firmographic data, and engagement signals into a live lead score — through one OpenAI-compatible endpoint, billed in rupees on a monthly GST invoice.

Why score leads with an LLM instead of fixed rules

Point-based scoring (job title, company size, a form field) is easy to game and blind to context. A language model reads a lead's actual behavior — page views, email replies, support questions — and reasons about buying intent the way an experienced sales rep would, without the rep's fatigue or bias.

Signals worth feeding the model:

• Recent activity (pricing page visits, demo requests, downloaded resources)
• Firmographics (company size, industry, funding stage)
• Engagement history (email opens/replies, webinar attendance)
• Explicit intent (form answers, chat transcripts)

Models for lead scoring

Model	Input ₹/1M	Output ₹/1M	Fit
Qwen3 235B-A22B	₹9.07	₹55.44	High-volume, budget scoring
Claude Sonnet	₹201.6	₹1008	Reads nuance in free-text notes
GPT-4.1	₹201.6	₹806.4	Structured reasoning, explains the score

Prices per 1M tokens — see live rates at /models.

Score a lead in one call

from openai import OpenAI

client = OpenAI(
    api_key="ub-gw-...",
    base_url="https://api.unoblox.ai/v1"
)

lead = {
    "name": "Rajesh Kumar",
    "company": "TechCorp India",
    "recent_activity": "Viewed pricing page 3x, downloaded case study, replied to email",
    "company_size": "50-200 employees",
    "industry": "SaaS"
}

response = client.chat.completions.create(
    model="qwen/qwen-3-max",
    messages=[{
        "role": "user",
        "content": "Score this lead 1-10 on likelihood to close this quarter. Return JSON with score and reason.\n" + str(lead)
    }]
)

print(response.choices[0].message.content)

A simple scoring rubric

Score	Meaning	Action
8–10	Hot — active buying signals	Route to senior rep, same day
4–7	Warm — engaged, not urgent	Nurture sequence, weekly follow-up
1–3	Cold — low or no signal	Automated drip, no rep time

Where teams use this

• B2B SaaS pipelines, to stop reps chasing dead leads
• Real estate, ranking site-visit requests by budget fit
• Insurance renewal follow-up prioritization
• Recruitment, ranking inbound candidate interest

Run scoring nightly on your full pipeline, or in real time the moment a lead crosses a trigger event (pricing page visit, demo request). A typical score call runs 150–400 tokens — at Qwen3 235B-A22B rates that's well under ₹0.05 per lead, so scoring 10,000 leads a month stays under ₹150.

Frequently asked questions

Q: Do I need CRM integration to start? A: No. Feed the model any structured record — an email thread, activity log, or form data — and it returns a score. CRM integration (Salesforce, HubSpot, Zoho) just automates the handoff.

Q: How do I validate the score is actually predictive? A: Back-test against closed-won and closed-lost deals from the last two quarters. Check whether high scores correlate with wins, then tune your prompt's weighting.

Q: Does this replace my sales team's judgment? A: No — it triages. Reps still make the call; the model just stops them spending equal time on every lead regardless of quality.

Q: Can I score leads with no purchase history, for cold outbound? A: Yes. Use firmographic fit (company size, industry, tech stack) plus any reply or engagement signal from outreach. Scores get sharper as behavioral data accumulates.

Q: What if a rep disagrees with the AI score? A: Log the override and the reason. Feed a sample of overrides back into your prompt as examples — the score should improve as you refine what "good fit" means for your business.

Q: Where is my lead data processed? A: Requests are billed and logged through unoblox's India-based gateway; each model you choose has its own processing location, listed on that model's /models page.

Get started in rupees → https://unoblox.ai/sign-in