AI scoring

AI qualification tells you who to attend to first: it analyzes a lead’s conversation and returns a 1-to-10 grade, their buying intent and a breakdown that explains the score.

How to qualify a lead

1. Open the lead’s record (/inbox/leads/:id).
2. Open the AI Analysis tab.
3. Click Qualify with AI.

In a few seconds you’ll see:

• Overall grade from 1 to 10.
• Intent: buy, evaluate, browsing or not qualified.
• One-sentence summary.
• Breakdown by criterion, with each criterion scored from 0 to 100 and visual bars.
• Evidence: chips that point to the actual messages each criterion is based on. Clicking them jumps to the conversation and highlights the quoted message.

At the bottom you’ll see the model used and the date.

It’s real AI, and honest

Note

Qualification uses a real language model. It’s not a formula disguised as AI.

Three design guarantees worth knowing:

• Quotes are not made up. The AI can only quote messages that genuinely exist in the conversation. If it tried to reference something nonexistent, it’s discarded. You’ll never see fabricated “evidence.”
• The overall grade is stable. The final 1-to-10 is calculated by weighting the criteria of your rubric; it’s not a standalone number the model tosses out. This makes it more reproducible than asking an AI to “give it a grade.”
• The model doesn’t add criteria. Only the criteria of your rubric are scored. If the AI omits one, that criterion stays at 0 marked as “not tested,” not as made up.

Like any AI, the text and the grade may vary slightly between two runs; what doesn’t change is that the evidence always points to real messages.

The rubric: what it’s scored against

The rubric is the template of criteria against which the AI judges each lead. Each criterion has a weight. By default a BANT-style scheme is used (Budget, Authority, Need, Timeline), and you can define custom rubrics per pipeline (for example, one for real estate and another for a clinic).

Challenging the grade

If you disagree with the score, you can write a comment and click Challenge grade: the AI re-scores taking your observation into account. That text is used only to re-evaluate; it’s never saved as evidence (the evidence is still the real messages).

AI consumption

Each qualification consumes AI credits from your plan. If you run out of the month’s allowance, qualification stops in a controlled way:

Caution

When you hit the credit cap, AI qualification stops running, but the lead is not lost: the contact is kept and you can keep replying to it by hand. Only AI scoring is paused.

Check how many credits each plan has in Plans.

Note

AI qualification only works against the real (cloud) environment. In local test mode it is not available.