Forty inbound leads arrived this week. Six are worth a call, eleven are worth a reply, and the rest will quietly eat your Tuesday if you let them.
Every page on this subject frames qualification as scoring leads in. The useful frame is scoring them out, because disqualification is the only qualification decision that gives you time back. If the question is how fast you reply, that is speed to lead; if it is where qualification fits in a prospecting day, that is AI prompts for SDRs. This page owns the qualify-or-drop call and the handoff after it.
The decision is what to drop, not what to work
A rep who works every inbound lead has no capacity left for the six that matter, and the six do not announce themselves. The first prompt forces a ranking, because a scoring model with no floor approves everything.
Here are this week's inbound leads with their form data and any activity. Rank them into three groups: work this week, hold with a reason to revisit, and drop. Force at least half into hold or drop, and for every lead name the single piece of evidence driving its placement.
The forced split is deliberate. Asked to score leads without a constraint, a model finds something encouraging in all forty, because encouragement is what the form data was collected for.
Prompts that score an inbound lead against evidence rather than enthusiasm
A demo request is an action, not a signal of fit. Job title is self-reported and company size is often wrong, so what is left is thinner than it looks.
Here is one inbound lead. Score it against our ideal customer profile, which I am giving you below. List what the evidence actually supports, what I am assuming, and what I would have to check. Do not infer seniority from job title alone.
Compare this lead against our last ten closed-won and ten closed-lost customers. Say which group it resembles and on which specific attributes, and name the attributes where I have no data either way.
The second is the honest version of lead scoring, and it only works if your profile is written down. Building it is covered in how to build an ideal customer profile with AI, with the sourcing side in Prospecting & Pipeline Creation.
Reading the form fill and the footprint for intent
Two leads with identical form data are not identical. One read three pricing pages at 11pm; the other downloaded a report and left.
Here is a lead's form submission and their recorded activity: pages, timing, repeat visits, content. Tell me what the pattern suggests about where they are in a buying process, and distinguish research behaviour from evaluation behaviour. Say plainly if the activity is too thin to read.
"Too thin to read" has to be an allowed answer, or you will get a confident narrative built on two page views. The broader signal work is in spotting buying signals.
Qualification questions for the first reply, not the first call
The expensive mistake is booking a call to qualify. Two questions in the first reply do the same job without spending the slot.
For this lead, write the two questions in my first reply that would most change whether I take a meeting. They must be answerable in one line each, must not feel like a screening form, and must not ask for budget.
Here is their one-line answer. Say what it tells me, what it conspicuously avoids, and whether it is enough to book the call or needs one more exchange.
The avoidance read is the valuable half: a prospect who answers the easy question and skips the other has told you something. Question design sits in Meeting Prep & Discovery.
Prompts that disqualify fast and politely
A bad disqualification costs you the referral and the second chance. A good one closes the loop so cleanly that they come back when the timing is right.
This lead is not a fit, for this reason. Write a short reply that is honest about the mismatch, points them to something genuinely useful, and leaves the door open without implying we will follow up. Four sentences.
Here are the leads I dropped this month and why. Group the reasons, and tell me which groups suggest a problem with our targeting or our form rather than with the leads.
That turns a month of rejections into a marketing conversation. Thirty leads dropped for one reason is not a lead quality problem, it is a form field nobody has fixed. Messaging craft is in Outreach & Messaging.
When the lead is right and the timing is not
The hold group is where pipeline is quietly lost. A lead marked "not now" with no date attached is a lead dropped with extra steps.
For each lead in my hold group, give the specific trigger or date that should bring it back, based on what they told me. Where no trigger exists, say so and propose the one question that would establish one.
Re-engagement later is its own motion, covered in the three-touch follow-up sequence.
Routing: prompts that write the handoff an AE will actually read
Write the handoff note for this qualified lead: what they asked for, what I confirmed, what I assumed, the two open questions, and the one thing the AE should not repeat because I have already asked it. Under 150 words, no narrative.
The "do not repeat" line is what makes an AE read the next one. The receiving end is covered in AI prompts for account executives, and the data hygiene behind routing in RevOps, Data & Analytics Support.
Take one prompt free, with no account, at /free-prompt. Pro is $12 per month or $89 per year, and lifetime access is $149 — See pricing.
Why a model given your form data will qualify everything
Hand a model forty form fills and ask which are good, and it finds a reason in each. That is not a flaw in the model; it is what the data is for. Forms are built to capture interest, so every row holds evidence of interest and almost nothing arguing against a lead.
It also cannot see what disqualifies people in practice: that this logo is already a customer under another name, that their procurement process killed a deal last year, that the person filling the form has no budget and never will. None of that is in the row.
So the constraints matter more here than anywhere else. Force a floor into the ranking, demand the evidence behind each placement, and allow "too thin to read" as an answer. Without the floor it is a machine for making you work every lead.
The rest of the catalogue, by role and by stage, is mapped in AI prompts for sales. Browse the library → or read the full taxonomy.
Frequently asked questions
What are good ChatGPT prompts for qualifying inbound leads?
The ones with a floor built in. Four earn their place: a forced ranking that pushes at least half the week's leads into hold or drop with the evidence behind each, a profile-comparison prompt against your last ten wins and losses, a two-question prompt for the first reply, and a handoff prompt that names what the AE should not ask again.
How do I use AI to score a lead?
Give it your written ideal customer profile and one lead, and require three separate outputs: what the evidence supports, what you are assuming, and what you would need to check. Tell it not to infer seniority from job title. Without the profile in the prompt, the score is just a summary of the form.
Can AI decide which leads to work?
No, and the way it fails is specific: handed form data alone it will approve almost every lead, because forms are built to capture interest. It is good at ranking against criteria you supply and at saying which evidence is thin. The judgement about capacity, and which six leads get your week, stays yours.
How do I disqualify a lead without burning the relationship?
Be honest about the mismatch in four sentences, point them somewhere genuinely useful, and leave the door open without promising follow-up. Then group your month's drop reasons: if thirty leads were dropped for the same reason, that is a targeting or form problem to raise with marketing, not a lead quality problem.
What data does AI need to qualify a lead?
The form submission, any recorded activity with timing and repeat visits, and your ideal customer profile written down. Activity patterns are what separate research behaviour from evaluation behaviour. What it will not have is whether the account is already a customer under another name, or how their procurement behaved last year.