When we reviewed real calls scored by our first version of AI lead scoring, we found a problem every sales team will recognise: 11 of 16 longer calls were marked hot, and none were cold. Pricing questions were read as buying intent. Existing customers calling for support were scored as new leads. Callers who said "abhi nahi" still came out warm.
If everything is hot, your team can't tell who to call first — which defeats the point. So we rebuilt the scoring to be evidence-based: it judges only what the caller said and did, writes down the proof before it chooses a label, and adds a new label for calls that were never sales opportunities.
Key Takeaway
ExoCall labels every call Hot, Warm, Cold, Callback Requested, Not Interested, or Not a lead, with a 1–5 score to match. Hot requires a concrete commitment on the call — not just interest. Every label comes with the deciding evidence and a quote of the caller's own words. On 36 hand-checked real calls, the new scoring got 31 right and found all 5 hot leads.
The Six Lead Labels, Explained
| Label | What it means | Score |
|---|---|---|
| 🔥 Hot | A clear need and a concrete commitment made on this call. | 4–5 |
| ♨️ Warm | A clear need and genuinely engaged — asked about pricing, features, setup — but no commitment yet. | 3 |
| ❄️ Cold | Vague or no need, just exploring, a real blocker, or declined the next step. | 2 |
| 📅 Callback Requested | Explicitly asked to be called back later. ExoCall schedules the callback. | 2–4 |
| ✕ Not Interested | Explicitly said they don't want it. | 1 |
| 🛠 Not a lead | A paying customer wanting support, a wrong number, a vendor, or a job seeker. | 1 |
What Makes a Lead Hot?
A hot lead is a caller with a clear need who made a concrete commitment on the call. Interest is not a commitment. Asking "charges kya hai?" is not a commitment. Any one of these is:
- Signed up, created an account or started a trial during the call — always hot.
- Chose a specific plan and asked how to start, buy or pay.
- Asked for an invoice, a payment link or UPI details.
- Said they want to start today or right now — and asked how.
- Agreed a specific day and time for the next step, like a demo or a site visit.
When the evidence sits between two labels, the scoring chooses the lower one. A slightly pessimistic hot list your team trusts is worth far more than an optimistic one they learn to ignore.
Why "Not a Lead" Matters
Many calls to a business aren't sales at all. An existing customer asking how to reset a setting. A wrong number. A vendor pitching their services to you. Scoring those as cold leads drags down your campaign numbers and clutters your lead list.
Not a lead separates them out — but carefully. Only someone who has already paid counts as a support call. A free-trial user asking for setup help is still deciding whether to buy, so they stay a lead (usually warm). And a paying customer who asks about upgrading is a lead again.
Every Label Comes With Proof
Expand any call in Call Logs and you'll see Why This Label: the deciding evidence in one or two sentences, with the caller's own words quoted in the language they spoke.
The same reason is in your CSV export and in the call.analyzed webhook, so your CRM gets the evidence along with the label. If you ever disagree with a label, you can see exactly why it was chosen in seconds.
How the Scoring Works Under the Hood
The biggest lesson from testing: an AI model is far better at answering one clear yes-or-no question than at weighing a whole rulebook at once. So after each call, the AI listens to the recording and fills in a fact checklist — each answer backed by a quote:
- Is the caller already a paying customer?
- Did they sign up or start a trial on the call?
- Did they choose a specific plan? Separately: did they ask how to start?
- Do they want to start now or today?
- Did they ask how to pay?
- Did they agree a specific day or time for the next step?
- Did they decline or postpone the next step?
- Is there a real blocker — a price clearly beyond them, a provider they won't switch from, or a product that doesn't fit?
Then fixed rules — not the AI's mood — apply the label. A hot label with no quoted commitment is lowered to warm. A warm lead who declined the next step becomes cold. When a rule changes the AI's first choice, a note is added to the reason, so nothing changes silently.
Two details matter a lot in practice. Wanting to speak to a human, or being annoyed with the call itself, is not a blocker — it goes into objections, and the lead keeps its label if they still want the product. And background noise — a TV, other people in the room — is ignored.
Tested on Real Calls
We tested each version against 36 real calls — a 16-call set we tuned on and a fresh 20-call set we didn't — labelled by hand after listening to every recording:
First version
16 / 36
correct labels
—
Evidence-based
24 / 36
correct labels
1 of 5 hot leads found
Fact checklist (live)
31 / 36
correct labels
All 5 hot leads found
Bottom Line
Lead scoring is only useful if your team trusts it. ExoCall's AI now scores every call on what the caller actually said, shows its evidence, and keeps support calls out of your pipeline — so "hot" means call them now, and "cold" means you can safely wait.
Live now for all ExoCall workspaces
Every newly analysed call uses the new scoring. Filter by label on the Leads page and in Call Logs — including the new Not a Lead filter.
Frequently Asked Questions
What is a hot lead in AI call scoring?
In ExoCall, a hot lead is a caller with a clear need who made a concrete commitment on the call — for example signed up or started a trial during the call, chose a specific plan and asked how to start or pay, asked for an invoice or payment link, said they want to start today and asked how, or agreed a specific day and time for the next step. Interest, questions or asking the price alone are never hot.
What does "Not a lead" mean?
"Not a lead" is for calls that are not sales opportunities: a paying customer calling for support or setup help with no new purchase discussed, a wrong number, a vendor pitching to you, or a job seeker. Anyone who hasn't paid yet — including free-trial users asking for help — is still a lead. A paying customer who wants to upgrade or buy more is a lead too.
Why does every ExoCall lead label come with a quote?
Because the label should follow from evidence, not the other way round. The AI writes down the deciding evidence — including a short quote of the caller's own words in the language they spoke — before choosing the label. You see it as "Why This Label" in Call Logs, in the CSV export, and in webhooks, so you can check any label in seconds.
How accurate is ExoCall's AI lead scoring?
On 36 real calls hand-labelled by listening to the recordings, the current scoring got 31 labels right and found all 5 hot leads. The previous version got 24 right and found 1 of the 5 hot leads; the original version got 16 right.
What is the difference between Cold and Not Interested?
Cold means low intent: a vague need, just exploring, a real blocker like a price clearly beyond them, or declining the next step ("abhi nahi"). Not Interested means the caller explicitly said they don't want it. Callback Requested is for callers who asked to be called back at a later time — ExoCall schedules that callback automatically.