Under the surface

How it works

A high-level view of what sits behind every outcome we deliver. No product names. Just the structure.

Where your data moves

Your systems stay where they are. Afferentic hosts the connectors and orchestration on our own dedicated infrastructure, isolated per client. Data flows between your platforms through this layer — we build and operate the connections.

Your systems
CRM, accounting, project management, email — whatever you use
API
Hosted by Afferentic · Isolated per client
Connector layer
One connector per system. Official APIs. Scoped permissions. Revocable.
Orchestration
Workflow engine
Decides what happens, in what order. Learns your patterns. Routes to managed processing only when needed.
Separate schemas, queues, and processing per client. Your data stays completely isolated.

For work that requires external reasoning or judgement, data is tokenised (PII stripped) before reaching the managed processing layer. The processing engine never sees real names, amounts, or identifiers.

Pattern detection and anomaly monitoring run at the orchestration layer — in your dedicated tenant, using your real data through the API connections you authorised. This is how the system learns your business without sharing data between clients.

The four layers

Every outcome we build uses the same architecture. Four layers, each with a distinct job.

Your systems
You own these
Accounting
CRM
Project mgmt
Email
Spreadsheets
+ others
Official APIs · OAuth2 · Scoped permissions
Connectors
We build these
One per system
Read / write as agreed
Isolated from each other
Independently revocable
Structured data · Event triggers · State management
Orchestration
The workflow engine
When X happens, do Y
Branching logic
Human approval gates
Retry and error handling
Only when judgement is needed · PII tokenised first
Processing
Managed tier only
Pattern recognition
Classification
Document extraction
Anomaly detection

The outcome lifecycle

Every outcome follows this cycle. Not just once — continuously, for as long as it runs.

1
Trigger
A defined event fires the workflow
2
Execute
The task runs — connecting, processing, or routing
3
Verify
The output is checked against expected results
4
Monitor
Health, drift, and exceptions watched continuously
5
Remediate
Issues fixed or escalated — without being asked

This is the "operational ownership" from the premise. Not a script that runs once. A system that owns the outcome end to end — the same way a responsible person would.

What stays with you

The playbook
Every outcome is documented: what it connects to, what permissions it uses, how the workflow runs, what happens if it fails. Your IT team can read it. Your next provider can maintain it.
The audit trail
Every action logged. What triggered it, what data was accessed, what the decision was, what the result was. Immutable. Exportable. Yours.
The knowledge
Processes documented from how they actually work, not how someone thinks they work. Edge cases captured. Institutional knowledge that no longer lives in one person's head.
Standard APIs
No proprietary protocols. No lock-in. Everything built on the same interfaces your systems already expose. Any competent developer can maintain it.
Next: Security and governance →