GetOut Commercial AutomationFrom a sports facility to a reviewed commercial action.
I structure and implement a n8n prospecting pipeline: discovery, operator identity, ICP qualification, enrichment, scoring, M1–M4 drafts and human review. It connects to the custom CRM without merging the two systems.
I structured discovery around Data ES and SerpAPI. An equipment record or a website is a starting point, not proof that the organisation operates the venue.
Inputs, outputs & controls
Input
Registry record / search result
Output
A sourced candidate and a job in the processing queue.
Protection
Deduplication and traceable source references; no contact is authorised here.
I separated the facility, the legal organisation and its official web identity. Directories, web agencies and suppliers must not be mistaken for operators.
Inputs, outputs & controls
Input
Candidate + official/site evidence
Output
Identity evidence linked to the organisation.
Protection
Ambiguous or contradictory identities go to human review.
I formalised ICP qualification with eligible, rejected and manual-review outcomes. Fit and intent scores order work; they never turn an ineligible organisation into an eligible one.
Inputs, outputs & controls
Input
Identity and sports-facility evidence
Output
An explicit qualification decision with reasons.
Protection
Missing evidence blocks progression; no favourable default.
I connected site exploration, facility facts and contact discovery. Facts retain their origin and missing information remains visible rather than being invented.
Inputs, outputs & controls
Input
Qualified organisation + accessible sources
Output
Sports, facilities, contacts and provenance.
Protection
Conflicting facts or uncertain contacts require review.
I connected fit and intent signals to commercial prioritisation, separately from the eligibility decision.
Inputs, outputs & controls
Input
Eligible record and enrichment signals
Output
A prioritised record, with uncertainty retained.
Protection
A high score cannot bypass exclusions or review.
I designed the M1–M4 sequence and its context rules. The LLM personalises a limited part of the message; a deterministic fallback covers weak context or provider failure.
Inputs, outputs & controls
Input
Qualified facts and an appropriate contact
Output
Drafts, not permission to send.
Protection
No unsupported claim; sequence state and previous steps remain controlling.
I specified review, final send gates and recovery paths. Deduplication, opt-out, blacklist, limits, retries and idempotence protect the process; an uncertain delivery is not blindly retried.
Inputs, outputs & controls
Input
Drafts and current sequence state
Output
A review decision or a blocked next action.
Protection
Some approval-safety improvements remained local in the latest report. No autonomous campaign is claimed.
I designed the link between the workflow lead and the commercial record. The CRM reads pipeline data through FDW; reviewed actions use dedicated Edge/RPC paths.
Inputs, outputs & controls
Input
Pipeline identity, contacts, drafts and status
Output
A manually confirmed link and a commercial next action.
Protection
Separate databases; no general bidirectional synchronisation.
The history documents lead generation, drafts and past sends. The latest safety changes have their own test and release status; neither leads nor drafts prove meetings or revenue.
Inputs, outputs & controls
Input
Execution history and dated test/release reports
Output
An inspectable maturity assessment.
Protection
No production run or message is triggered by this portfolio.
The result and the next step.
Sourcing, drafts and historical sends are documented; recent safety changes have separate local/deployment states. No current autonomous campaign or commercial uplift is claimed. The illustrated prospect is fictional.