ORCHESTRATION
Sixteen workflows
Coordinated through n8n
BUSINESS DEVELOPMENT · AI WORKFLOW AUTOMATION · INTERNAL TOOLS
Building an evidence-led account intelligence workflow.
We built our own account-intelligence system to source companies, analyse their websites, enrich the right people, prepare personalised outreach and coordinate manual LinkedIn execution without turning the process into a black-box sending bot.
- Owner
- Tested Works
- Sector
- Business development
- Engagement
- Internal operating system
- Status
- Live and evolving
DATA
One source of truth
NocoDB and PostgreSQL
OUTREACH
Human-controlled
The system prepares and reconciles; the operator acts
PROJECT SNAPSHOT
The engagement in three parts.
01
Situation
Useful prospects were scattered across directories, websites and professional networks. Generic databases could provide contact details but not enough evidence to explain why a company or person was relevant.
02
What we delivered
A self-hosted pipeline for scraping, site analysis, company scoring, employee enrichment, people qualification, outreach writing, task reservation, status reconciliation and operator review.
03
Our responsibility
We own the workflow definition, orchestration, shared source of truth, evidence retention and continued operation.
What we operate
One live internal workflow connecting difficult lead sources, company evidence, people enrichment, AI qualification and human-controlled outreach.
- Sixteen n8n workflows share a PostgreSQL-backed source of truth and explicit step status.
- Scraping, enrichment, language models, Slack and a Chrome operator console operate as coordinated components rather than isolated automations.
ORCHESTRATION ARCHITECTURE
Evidence stays beside every recommendation and action.
Orchestration coordinates the work, but shared state and a visible human action boundary make the internal system controllable.
Sources
- Named data sources
- Website evidence
Qualification
- Company classification
- People discovery
Preparation
- Evidence briefs
- Message preparation
Human action
- Review, approval and outreach
State and reconciliation
Shared state, evidence retention, reservations and operator-confirmed relationship updates support every stage.
SELECTED SYSTEM EVIDENCE
From difficult sources to a controlled outreach action.
01
A source-specific company universe
Crawlee and Playwright services collect companies from named directories incrementally, with source-aware deduplication, repeat handling and crawl status. A shallow website service then gathers the pages most useful for qualification.
Why it mattered
The system begins with defensible company sources instead of an undifferentiated contact export.
02
Evidence-backed company and people qualification
Language-model tasks score company fit from the website evidence. Employee records are enriched and ranked across buyer, influencer, champion, role-fit, workflow evidence and disqualification dimensions.
Why it mattered
The selected person is connected to a specific account hypothesis rather than only matching a seniority filter.
ENGINEERING DECISIONS
Keeping evidence, state and human control visible.
Enrich people only after the company qualifies
Directory and website evidence is cheaper than contact enrichment. The pipeline discards weak companies before paying for deeper people data and language-model work.
Workflow economics
Keep evidence beside every recommendation
Company and person scores include the reason, observed signal and proposed angle. Operators can understand why the system prioritised an account instead of receiving an opaque rank.
AI trust
Separate preparation from execution
The automation reserves and prepares actions, while the operator performs them in the external platform. Read-only page capture and explicit confirmation keep the workflow useful without turning it into a sending bot.
Human control
RESPONSIBILITY
Why we built and operate it
Repeated internal need
- Research consistency
- Evidence retention
- Visible operator control
End-to-end ownership
- Workflow and data model
- Orchestration
- Operation and improvement
Live internal operating workflow · continuously evolved
SERVICES USED
SELECTED TECHNOLOGY
n8nNocoDBPostgreSQLCrawleePlaywrightApolloApifyOpenAISlackChrome Extension APIs
RELEVANT EXPERIENCE
Relevant if your situation includes…
You need to join several imperfect data sources into one operating workflow.
You need AI recommendations to retain the evidence behind them.
You need automation to prepare and coordinate work while a person controls the consequential action.
RELATED WORK
Two adjacent examples.
CONFIDENTIAL PARTNERSHIP PLATFORM
Automating a mobile-only partnership workflow without an API.
Three platform integrations with custom visual models and verified human-approved actions.
View case study
CONFIDENTIAL HOSPITALITY OPERATOR
Making multi-site data pipelines recoverable and easier to operate.
APIs, databases, managed files and browser-only reports coordinated in one Dagster estate.
View case studySTART A CONVERSATION
Is your team the integration layer between databases, websites and manual decisions?
We can map the workflow, connect the evidence and build the internal tool that makes the process visible and controllable.