AI lead scoring can turn a CSV of inbound inquiries into a list with an explicit priority signal. In Magic CSV, Jev evaluates each message against the endpoints you define and adds a 0–100 score. The score describes how the supplied text fits your rubric; it is not a probability that the lead will become a customer.
What the result looks like
Illustrative example with synthetic data and authored answers, not a recorded model run. Your results will depend on the source data and instructions.
Scroll tables sideways to see every column.
| # | Company | Inquiry |
|---|---|---|
| 1 | Example Studio | Please send pricing for 12 seats. We want to start this week. |
| 2 | Sample Works | Researching options for next year. Could I read a case study? |
| 3 | Demo Labs | Please remove me from your mailing list. |
| # | Company | Inquiry | Purchase intent |
|---|---|---|---|
| 1 | Example Studio | Please send pricing for 12 seats. We want to start this week. | 90 |
| 2 | Sample Works | Researching options for next year. Could I read a case study? | 25 |
| 3 | Demo Labs | Please remove me from your mailing list. | 0 |
Try it, step by step
Start with evidence in the file
Import a CSV or Excel file containing Company and Inquiry. Include stated timing or requested seats if you have them. A list containing only names and email addresses does not provide evidence of purchase intent. Use the sample to try the setup before uploading a real export.
Define Purchase intent
Choose AI actions → Score. Name the column Purchase intent. For 0 means, enter “No interest in buying; unsubscribe requests or unrelated messages.” For 100 means, enter “Explicit purchase request with a near-term start date.” Under Read these columns, select Inquiry.
Constrain the judgment
In More guidance → Extra instructions, enter: “Use only evidence stated in the inquiry. Do not infer budget, authority, or urgency from a company name or email address. Research without a buying timeline should score below an explicit request to purchase soon.” Run the column.
Review, then prioritize
Check high and low scores against the messages, and inspect uncertain cells. Correct individual scores when needed. Export the reviewed file and sort Purchase intent from highest to lowest in Excel or Google Sheets. Keep the original inquiry visible when deciding how to follow up.
Separate intent from customer fit
A company can fit your target audience and still have no current interest in buying. Conversely, a small company outside your usual segment might be asking to purchase today. Combining those ideas into one unexplained score makes it difficult to understand why a row ranked highly.
Create a separate Fit score if the file contains the business context needed to judge it. Define the customer profile in Extra instructions, and select the relevant supplied fields. If team size or use case is missing, gather that information rather than asking the model to invent it. Keep unsubscribe requests separate from outreach candidates.
Test the rubric on your own inquiries
Choose examples your team would prioritize differently: an explicit pricing request, a student research question, a renewal question, a vague demo request, and an opt-out. Decide the intended order before running the model. Compare that order with the scores. If the endpoints leave too much room for interpretation, remove the test score column and create a replacement with revised definitions.
A score of 90 does not mean a 90% conversion rate. Nor does a two-point difference establish a meaningful difference between leads. Treat the result as a review aid tied to your instructions. Validate it against your workflow before relying on it for a larger list.
What this workflow does and does not enrich
The new column adds a judgment about data already present. It does not visit company websites, find contact details, infer verified revenue, or check a CRM for past conversations. A sparse lead list needs additional source data before it can support a useful contextual score.
Magic CSV does not send outreach or sync scores into your CRM automatically. Export the checked rows and use your normal import and review process. Preserve a source identifier so each score can be traced to the original inquiry, and avoid sending irrelevant personal columns to the model.
Common questions
- Can I score leads from an Excel file?
- Yes. Import CSV or Excel, add a Score column, and export either format. AI sheets support up to 1,000 rows, so begin with a small representative file to evaluate the rubric.
- Does Jev research each company?
- No. This workflow evaluates the source columns you select. Add verified company information to the file yourself if it is needed for a Fit score.
- Is the score a prediction of sales revenue?
- No. It measures alignment with your stated rubric using supplied text. It is not a conversion probability, forecast, or independently verified measure of lead quality.