Fast follow ups fail when lead handoffs leak context

You don’t have a lead problem. You have a handoff problem disguised as “we’ll follow up soon,” where context leaks out of forms, calendars, and inbox threads until the rep is guessing, the prospect is waiting, and the CRM gets updated after the outcome is already decided. Speed doesn’t fix that. Structure does. This playbook builds a working inbound triage system that captures intent, qualifies it, and routes it with receipts using Webflow, n8n, HubSpot CRM, and Perplexity. No extra dashboards. Just fewer dead leads. Outcome: every inbound submission gets enriched, scored, and assigned in under two minutes, with a clear audit trail for why it went where it went. Webflow is the front door. The form is not “name/email/message.” It’s a structured intake: use-case dropdown, timeframe, budget range, team size, and a single free-text field that forces specificity. You’re designing for routing, not politeness. n8n is the traffic controller. When Webflow fires the webhook, n8n normalizes fields, dedupes by email + domain, and checks HubSpot for an existing contact/company before it creates anything. Then it enriches: Perplexity pulls a quick company snapshot (what they sell, who they sell to, recent funding/scale signals) and returns citations you can keep for compliance and later review. No mystery scoring. HubSpot becomes the system of record, not the system of regret. n8n writes: lifecycle stage, lead source, enrichment notes, an explicit score breakdown, and an owner assignment based on rules you can change without retraining the team. It also opens a task with a prewritten first-touch email draft that references the intake plus one enrichment detail. The hard rule: if a rep overrides routing, they must choose a reason code. That’s how the workflow improves instead of drifting back into folklore.

Maya runs growth at a 25-person B2B SaaS. Monday, 9:12 a.m., a Webflow form comes in: “Looking at onboarding analytics. Need to fix drop-off. 2–3 months.” That’s not a lead. It’s an intent packet. n8n catches the webhook, normalizes “2-3 months” into a timeframe bucket, pulls the email domain, and checks HubSpot. Existing contact? Yes. Old deal lost six months ago. Different champion though. Here’s where it usually breaks. Someone “improves” the form by adding a friendly open text field and removing budget. Response rates go up. Routing gets worse. Now every submission looks like “Interested in learning more.” n8n can’t score ambiguity. The rep gets an alert with no spine. They wait. Prospect waits longer. Everyone blames speed. In the working version, n8n enriches via Perplexity. Not a novella. Three bullets and citations: what the company sells, who they sell to, and a scale signal. It returns that they’re hiring two product analysts and recently migrated to Snowflake. Maya doesn’t need perfect truth. She needs a reasoned guess with receipts. HubSpot gets updated before the first reply. Lifecycle stage set. Lead source stamped. Enrichment note pasted with citations. Score breakdown written as properties: Fit 3/5 (ICP adjacent), Intent 4/5 (specific use case + timeframe), Urgency 3/5, Authority unknown. Owner assignment routes to the rep covering “data products” accounts, not whoever happened to be online. Friction still shows up. The rep overrides the owner because “I know someone there.” They have to pick a reason code. Relationship. Territory exception. Strategic account. Annoying? Yes. Useful later? Also yes. Two minutes later a task appears with a first-touch draft referencing the drop-off use case plus the Snowflake migration. It reads human without being improvised. And the uncomfortable question: if your best leads require a rep to ignore the workflow to get handled correctly, is that a team problem or a system design problem?

Here’s the part people skip: getting this to run inside a real company isn’t a Zapier demo, it’s an operating change. The tech is the easy half. The hard half is making the workflow the path of least resistance for reps and the least risky thing for ops. Start by treating the form as a contract with sales. If sales won’t commit to routing off timeframe + use case + budget band, don’t pretend. Bring two reps into the form design review and make them sign off on the scoring inputs. Not “looks good,” but “I will accept assignments based on these fields.” Then implement in thin slices. Week one: just dedupe + HubSpot lookup + write structured properties. No enrichment, no scoring, no fancy owner rules. You’re proving you can keep data clean and avoid duplicate contacts. Week two: add scoring, but keep it dumb and explicit. Fit, Intent, Urgency, Authority. If someone proposes a weighted model, ask them what decision it will change on Monday. Week three: add Perplexity enrichment, but put it behind a guardrail: only enrich if the score crosses a threshold or if the domain is new. Otherwise you’ll waste tokens on tire-kickers and train the team to ignore the notes. And store the citations in HubSpot as a first-class property, not hidden in an activity log. The override reason code is where this either becomes a learning system or a compliance annoyance. Make it useful for reps: if they override for “relationship,” let that auto-create a note prompting them to name the relationship and next step. If they pick “territory exception,” route it to ops for review. No ops review, no behavior change. Finally, do a weekly 20-minute routing retro. Pull ten overrides, ten fastest-won, ten fastest-lost. Adjust one rule, not ten. If you can’t point to a specific workflow change that came from real override data, you’re not building structure. You’re just building a prettier way to drift.