Skip to content
AgentSDR
Live demo

AI lead qualification

AI lead qualification from what leads say

AgentSDR reads each reply with AI and files the lead into your stages with a reason and a confidence. When it is unsure it holds the call for a person. It qualifies from conversations, not from a score on a cold list.

agentsdr.your-company.com/crm/actions

Action required

Every reply the AI has classified, with the answer it drafted. Approve, edit or reclassify.

In the queue

23

replies and follow-ups due

Drafts to review

11

written by your model

Follow-ups due

4

past their due date

Errors

0

nothing failed today

Queue

Newest reply first. Hover a reply or a draft to read it whole.

The short version

What CRM platforms support AI-powered customer scoring?

Large CRM suites and several sales tools offer predictive scoring: a model ranks contacts from profile and behaviour data before they say anything. AgentSDR does not. It is a CRM for outbound that qualifies a lead from the content of their reply, and it is clear about that boundary.

If your question is who to contact first, you need prospecting and enrichment, which AgentSDR supports with AI columns and enrichment providers in Tables and a lead database. If your question is how interested someone who replied actually is, that is what the CRM answers. Tools for the first question are compared in best AI CRMs.

How it works

How does AI qualify a lead in AgentSDR?

Six properties of the way the classification works.

  • From the reply

    The signal is what the person wrote, such as asking for a demo, a referral to a colleague or a no for now.
  • Your stages

    Interested, Customer, Not interested and Other, with subcategories and funnel stage numbers that you edit.
  • A reason you can read

    Each classification stores a short reason and a confidence on the record.
  • Held when unsure

    Low confidence, Customer, Other and backward moves wait as proposals for a person.
  • Calls count too

    Transcripts of calls placed from AgentSDR are classified, with a stricter rule before a call changes a stage.
  • Correctable

    Fix a misread in one click. A better category description or instruction fixes the next ones.

Two approaches

Predictive scoring or reply-based qualification?

Predictive scoring compared with reply-based qualification
Predictive scoringReply-based qualification (AgentSDR)
InputProfile, firmographic and behaviour dataThe words of the lead's reply or call transcript
OutputA number or tier per contactA stage and subcategory, a reason and a confidence
When it worksBefore the lead has said anythingAfter the lead responds
StrengthPrioritises a large cold listReads intent that no profile field holds, such as timing and objections
WeaknessCannot see what the person meantSays nothing about contacts who never reply
Who acts on itRouting rules or a repA sequence drafts the next step; a person decides the uncertain ones

Trust

When does the AI hold a qualification for review?

When AgentSDR applies a classification and when it holds it
SituationWhat happens
Confidence at or above the threshold, forward moveApplied. The lead moves and its sequence drafts the reply.
Confidence below the threshold (85% by default)Held as AI suggests, shown as Confirm classification.
Proposed CustomerHeld for review. A lead already a Customer is not moved out by the AI.
Proposed OtherHeld for review by default.
A move backward, or to an outcome with no stage numberHeld, because going back a step is a judgement about the deal.
An explicit request to stopApplied at once. The person is marked Do Not Contact on every channel.

Qualified leads then get a sequence and a draft, described in AI email replies and automated CRM. Everything lands in the inbox across channels; see omnichannel CRM.

Tuning

How do I make the qualification more accurate?

  • Write each subcategory’s description as a test the AI can apply, with an example of a reply that belongs and one that does not.
  • Add classification instructions for the cases your market produces, such as referrals to a colleague or requests to come back next quarter.
  • Correct wrong calls and then fix the wording behind them, instead of correcting each reply by hand.
  • Keep the Customer and Other reviews on until you trust the results.

FAQ

Questions, answered

What CRM platforms support AI-powered customer scoring?

Two kinds of tools use the phrase. Predictive scoring ranks contacts from firmographic and behaviour data, and several CRM suites offer versions of it. AgentSDR does something different: it qualifies a lead from the content of their reply, with a stage, a reason and a confidence. It does not produce a numeric score for a cold list.

What is AI lead qualification?

It is using AI to decide how interested or ready a lead is. In AgentSDR the AI reads each reply from email, LinkedIn, WhatsApp or a call transcript and files the lead as Interested, Customer, Not interested or Other, with a subcategory you define.

How does the AI decide, and can I see why?

It reads the reply with your category descriptions, your classification instructions and your knowledge. It records a category, a subcategory, a confidence and a short reason you can read on the record.

What happens when the AI is not sure?

The classification is held as a proposal. Below the confidence threshold (85% by default), or when it proposes Customer or Other, or a move backward, the record shows AI suggests and waits for you to accept or correct it.

Can it score leads before they reply?

No. AgentSDR does not rank a cold list. Qualification starts when a lead responds. For prospecting data and fit filtering you can enrich people with AI columns and providers such as Apollo in Tables, on your own keys, and decide the targeting yourself.

Can I define my own qualification stages?

Yes. You add, rename and re-describe subcategories, give each a funnel stage number and attach a sequence to it. The classifier works from your descriptions, so better wording gives better results.

Qualify from the conversation

Open the live demo to read classifications with their reasons, or self-host your own.