How to build
How to build an AI lead triage and routing system
Inbound leads decay fast, and most of the delay is not sales being slow — it is the gap between a form submission and anyone knowing it matters. Triage is a mechanical job: read it, work out who it is, decide if it is worth a call, route it with the context attached. Mechanical jobs done late are exactly what an agent is for.
What you’re building
A lead triage system reads every inbound enquiry, enriches it with what can be found about the company, scores it against the criteria you actually care about, and routes it to the right person with a short brief — rather than dropping a raw form into a shared inbox.
Who this is a good fit for
- Teams where inbound volume is high enough that good leads get buried
- Anyone routing by hand from a shared inbox or a form notification
- Sales leads who want consistent qualification rather than whoever-is-free
How to build it
Step 1
Describe what a good lead looks like to you
This is the whole system, so be concrete and be honest. Company size, industry, the phrases that signal intent, the ones that signal a student project. Vague criteria produce vague scoring — "enterprise-ish" is not a rule, "over 200 employees or asking about SSO" is.
Lead Triagelead-triageDescribe it
Build a lead triage agent. For every form submission, enrich the company from its domain, then score: high if 200+ employees or they mention SSO, security review or procurement; medium if they name a specific use case; low if it's a free-tier or student query. Route high to the AE channel within 5 minutes with a brief, medium to the queue, low to a nurture list.Build itStep 2
Connect the form, the enrichment and the destination
Three connections: where leads arrive, what you enrich them with, and where the routed lead lands. Web Search covers enrichment for most teams without another subscription; if you already have HubSpot or Salesforce, the lead can be written there with the score attached rather than living in a second place.
Lead Triagelead-triageConnect
Typeformread submissionsConnectedWeb Searchenrich by domainConnectedHubSpotcreate + update contactsConnectedSlackpost to #inboundConnectedStep 3
It builds the pipeline and the scoring rule
It wires the trigger, the enrichment step, your scoring logic and the routing branches. The score is not a black box — each lead carries the reason it was scored that way, which is what makes the thing correctable when it is wrong.
Lead TriageTrigger: new Typeform submissionEnrich: company size, industry, funding from domainScore: high / medium / low with stated reasonRoute: high → #inbound, medium → queue, low → nurtureWrite back to HubSpot with score + reasonStep 4
Correct it for a week, then leave it alone
Run it alongside your current process for a week and correct the scores it gets wrong. Each correction is a sentence — "agencies under 50 people should be medium, not low" — and the rule changes. After a week the disagreement rate is usually low enough to stop watching.
leads.trysomething.siteCompany Size Score Reason Routed to Northwind Retail 1,200 High Asked about SSO + procurement #inbound Kestrel Labs 45 Medium Named use case: eval pipeline Queue — (gmail.com) — Low Personal email, free tier query Nurture
What it does once it’s running
- Every inbound lead enriched from its domain — size, industry, signals
- Scored against your criteria, with the reason recorded
- Routed to the right channel or rep within minutes
- A short brief attached, so the first reply is not starting cold
- Written back to your CRM so the score lives with the record
Tools it connects to
What it won’t do
- Enrichment is only as good as the public footprint — stealth companies and personal emails will stay thin
- Scoring needs a week of correction before it is trustworthy; treat the first week as training, not production
- It routes and briefs; it does not replace the judgement call on how to actually run the deal
Start from a template instead
Each of these opens with the prompt already written. Edit it before you build.
Common questions
- How accurate is the scoring?
- It starts rough and improves quickly because you correct it in plain language. The honest framing is that the first week is calibration — run it alongside your existing process, fix what it gets wrong, and by the end the disagreement rate is usually low enough to trust.
- Does it work with my CRM?
- HubSpot, Salesforce, Pipedrive and Close can all be connected, so the lead and its score are written to the record you already work from rather than living somewhere new.
- What about leads it scores wrong?
- Every score carries the reason it was given, so a wrong one is diagnosable rather than mysterious. You correct the rule by describing the exception, and it applies from then on.
- How fast does routing happen?
- Within a few minutes of the form submission. The point of the system is closing the gap between arrival and someone knowing it matters, so speed is the main thing it is built for.
Related use case
AI agent for GTM & outboundFind, enrich, score, and draft outreach to your ideal customers automatically.
Build it in the next ten minutes
Start with a sentence. Leave with a working app on its own URL.