Lead intake for a small B2B agency
An internal build that turns scattered inbound inquiries into a single, enriched pipeline — so no lead waits on someone remembering to copy it into the CRM.
One
pipeline instead of three inboxes
Internal build — observed in our own system, not a client deployment
< 2 min
from inquiry to enriched CRM record
Internal build — observed in our own system, not a client deployment
Nothing
dropped without a logged reason
Internal build — observed in our own system, not a client deployment
Problem
Inquiries arrive through a website form, a shared inbox and the occasional LinkedIn message. Someone has to notice each one, judge whether it is worth pursuing, look up the company, and paste it into a CRM. It happens late, or not at all, and there is no record of what was ignored.
Before — the manual workflow
- 01
New inquiry lands in a shared inbox
- 02
A team member notices it (eventually)
- 03
They manually research the company and contact
- 04
They decide, informally, whether it is worth a reply
- 05
Some inquiries are copied into the CRM; some are lost
Intervention
A single intake system that watches every channel, enriches each inquiry with public company data, scores it against a written definition of a good-fit lead, and creates a CRM record with a suggested next step. A person still approves anything before outreach — the system removes the busywork, not the judgment.
Why this build exists
This is an internal build — a system we built and run ourselves rather than a paid client project — and it is labeled that way on purpose. It shows how we think about a common, unglamorous problem: inbound leads that depend on a person remembering to act.
The numbers below come from our own deployment, not from a client engagement. When we run this for a customer, the results section will say so and cite where each figure came from.
The system, in plain terms
Every channel feeds one intake step. Each inquiry is enriched and scored, then a CRM record is created with a recommended next step. A human approves outreach. The result is one legible pipeline instead of three places to check.
What this build has to get right
A system like this fails quietly rather than loudly, so the interesting work is in the instrumentation:
- Enrichment misses. Public data is incomplete for small companies. Anything the enricher cannot resolve goes to a human queue rather than being scored on missing fields.
- Score drift. A “good-fit lead” definition written in January is wrong by June. The score is reviewed against closed-won data, not left to run forever.
- Channel silence. If a shared inbox stops delivering, nothing errors — the pipeline just goes quiet. Each channel gets a heartbeat check so silence is itself an alert.