The agent runs a real-time qualification check on every form fill using the enriched contact and company data. Qualification criteria are configured by your team: required attributes (must be in these industries, must have this employee range), preferred attributes (scores higher if on this tech stack, scores higher if Series B+), and disqualifiers (student email domains, non-target geographies, company size below minimum threshold). Each criterion is weighted by predictive value — an attribute that's present in 80% of your closed-won deals is weighted higher than one present in 20%.

Once a lead is scored, the routing decision is immediate. Tier 1 (high ICP fit, 80+) routes to AE with urgent flag. Tier 2 (medium fit, 60–79) routes to SDR queue for same-day follow-up. Tier 3 (low fit, 40–59) enters a nurture sequence. Disqualified leads are suppressed from the sales queue with a reason code logged for reporting. The entire routing decision happens in under 60 seconds of form submission — with no human in the loop. Reps receive only the leads they should be working, already enriched and scored.

Not every qualification decision is clear-cut. A company that meets 3 of 4 required criteria but misses on company size by a narrow margin is an edge case — the automated system would disqualify it, but a rep might decide it's worth pursuing. The agent identifies these edge cases: leads that score close to tier boundaries, leads with conflicting signals (small company but large deal size indicated in the form), and leads from companies with prior CRM history that changes the context. These are flagged for a rep review queue — not automatically routed or suppressed.

A qualification model that isn't measured can drift — disqualifying leads that convert and qualifying leads that don't. The agent tracks the downstream outcome for every routed and suppressed lead: which Tier 1 leads became opportunities, which Tier 2 leads converted at rep follow-up, and which disqualified leads turned out to be misclassified. Model accuracy is reported weekly — including false positive rate (high-scored leads that didn't convert) and false negative rate (suppressed leads that would have converted). These inputs feed quarterly model recalibration.