A prompt that scores every inbound lead against the criteria your best customers actually share, says why it scored what it did, and routes the ones worth a rep's time — with the borderline cases left visible instead of quietly binned.
Not because they cannot tell good from bad, but because the sorting never happened before the calendar filled up.
Every team has an opinion about it and no record of it, so qualification happens as a vibe during a call that a rep makes fifteen minutes after the form fill. Without written criteria, the same lead gets graded A by sales and C by marketing, and nobody can say which was wrong — because there is no standard to have been wrong against.
Company size, role, budget, industry, use case — all of it is usually captured at signup and then ignored, because reading thirty submissions carefully takes longer than it is worth when you expect a few to matter. The consequence is that the whole queue is treated as one undifferentiated mass and handled at the speed of the least interesting lead in it.
The first responder wins a disproportionate share of inbound, and that is well established enough that everyone knows it. What it means in practice is that a lead who would have been a great customer often gets disqualified for being slow — not by the criteria, but by the queue they sat in behind leads that were never going to close.
Copy it once, paste it into Zaira, and it grades the whole inbound queue: pulls every new submission, scores it against your closed-won history rather than a generic rubric, writes the reasoning back to the contact, and hands the strong ones to a named owner.
Three stages, one goal. Every lead is graded within minutes of arriving, so first response stops being a race against whoever checked their phone first.
Because the scoring runs on arrival rather than on a rep finding a gap in the day, the response-time gap that decides who wins inbound largely stops existing. Nobody is disqualified for being third in the queue.
Step 1 derives the scoring criteria from 25 won and 25 lost deals rather than a framework designed for a different business. The result is a rubric you can defend to sales because it is your own history, not a vendor's template.
Anything scoring between the two thresholds goes to a review queue with its score and reasoning, instead of being quietly binned. That band is where most of the surprises are, and it is the one most systems delete before anyone ever looks.
The practical questions people ask before scoring their leads by machine.
It will, and the design assumes that rather than denying it — that is why Step 1 derives the rubric from your own closed-won and closed-lost history instead of a generic framework, and why borderline leads go to a review queue rather than to the bin. Start by running Steps 1 and 2 only and reading a week of scores against what your reps would have said. You are looking for systematic disagreement, not perfection, and if the band is wrong the fix is your context section.
No — Step 4 drafts only, and it is worth keeping that way for a while. First contact is where a misjudged assumption is most expensive and most visible to a prospect you wanted. The scoring and routing are where the time pressure actually lives, and those are worth running unattended; sending a personalised first message is a decision, and it stays yours until the evidence says otherwise.
As many as arrive — the work per lead is a record read and a written score, so the volume that matters is how many arrive, not how long each takes. Zaira charges per plan rather than per task, so a spike in submissions is not a separate bill. Because the prompt is built to run repeatedly against new arrivals, put it on a schedule instead of triggering it by hand and the scoring keeps pace with the queue on its own.
It replaces the screening that was standing in for qualification, not the conversation. A score built from a form fill and a company record is a good filter and a poor substitute for listening — which is why the highest-value use is the first question in Step 4, aimed at whichever unknown would most change the score. Reps get fewer calls and better ones, because the screening call no longer has to do both jobs.
Then the honest answer is that qualification on arrival will be weak, and Step 5 will tell you so — leads with too thin a record to score honestly are reported as a group rather than given a confident wrong number. Two fixes worth doing first: ask for two more qualifying fields in the form, since data you never captured cannot be scored, and check whether the company record alone carries enough. Adding one good question to a form is usually worth more than any scoring sophistication.
Zaira works through HubSpot's own official MCP server — the permissioned connection offered to approved AI partners, not a scraper or a browser rig. There is no shadow portal; reads and writes happen inside the access you authorized, and you can revoke it from your HubSpot settings at any time. Disqualified leads are marked, never deleted, so the record of what was rejected and why survives for whenever you want to review the threshold.