AIDA — AI Decision & Activation Platform
Data to Insight. Insight to Decision. Decision to Action.
A multi-tenant control layer on top of existing tools. Instead of using separate dashboards, monitoring tools and security alerts, AIDA brings the relevant facts together into one operational model — and builds, in a controlled way, toward safe automation.
What AIDA is — and isn't
AIDA is not just a dashboard, not just a chatbot and not just an automation tool. It's a controlled platform layer that helps organisations move from scattered signals to understandable, explainable and demonstrable action.
Overview
Organisations usually don't lack data, but coherence. Information is spread across tools, interpretation is manual and follow-up depends on individual knowledge. AIDA is the central platform layer where data, insight, judgment, follow-up and controlled automation come together in one coherent system.
It collects signals from different systems, makes them comparable, keeps context and history, detects anomalies, helps judge what matters, organises actions and records the outcome. The result is one traceable chain instead of separate tools.
The three questions AIDA answers
What's happening?
Trends, anomalies, alerts and summaries based on factual data — overview without forcing a conclusion.
What does it mean?
Classification, impact, severity, priority and recommendation. Claude adds interpretation, but only on grounded context.
What do we do now?
Tasks, playbooks, approvals and workflows. Automation comes later — and only within explicit guardrails.
The core principle
Decision ≠ Action
The layer that judges and advises may not change systems or send messages on its own. Execution always runs through the Action Engine, under permissions, policies, audit and — where needed — human approval. This prevents a faulty classification or a hallucinating language model from directly causing an external change — and keeps AIDA explainable, auditable and enterprise-ready.
The architecture in layers
AIDA Portal
Next.js/React workspace with dashboards, issues, decisions, actions, reports and an AI Copilot. Never talks to the database directly.
API & Access Layer
NestJS gateway with authentication, RBAC, tenant context, entitlements and audit. Every request stays within the right permissions and boundaries.
Data Sources
The sensory layer: Cloudflare (MVP), analytics, security tools, SEO scanners, CRM/ERP, logs and webhooks deliver raw facts.
Data Ingestion
Collectors, normalization, scheduling (BullMQ/Redis), secrets and quality control — the translator between source and Core.
Data Core — PostgreSQL
The single source of truth. Tenant Data Plane with Row-Level Security + an aggregated, anonymised Knowledge Plane.
Intelligence & Activation
Insight, Decision and Action engines with Claude as a grounded AI runtime — the platform's distinctive value.
Integrations & Output
Ticketing, ITSM/ServiceNow, email/Slack/Teams, webhooks and APIs — insights and actions land in existing workflows.
Operations & Reliability
Jobs & queue, event backbone, notification connectors, monitoring, backups/DR and secrets — AIDA runs reliably in production.
Control Plane (BSS)
Tenancy, partners, plans, entitlements, marketplace and FinOps/billing (via Lago) — per-customer differences are configuration, not a code fork.
Governance & Security
RLS, least-privilege RBAC, audit, policies/compliance and kill-switch/guardrails — woven through every layer, not a separate screen.
The five movements
Data
Sources deliver signals; collectors pull them in; the ingestion layer normalises and validates; metrics, events and findings land in the Tenant Data Plane.
Insight & judgment
Insight reads the Core and surfaces anomalies; Decision combines rules, context, policies, history and optional Claude interpretation.
Follow-up
An issue or decision is turned via Action into a task, playbook, workflow or — in a controlled way — an automated integration.
Learning
A completed action gets an outcome. Only anonymised, aggregated patterns flow to the Platform Knowledge Plane.
Control & scale
API, governance, RLS, RBAC, entitlements, partners, marketplace and FinOps decide who sees what and how AIDA scales safely and commercially.
Tech
Approach
The architecture is deliberately bigger than the first pilot: a working MVP core (Cloudflare → Core → Insight → Portal → light Decision with Claude → reporting/notification → simple action) that grows in a controlled way into a scalable ecosystem — without replacing the core.