This is the engine that thinks for all of them.
Qbiqal Cortex AI Engine is the purpose-built intelligence layer that powers every AI employee across all six Qbiqals. Four sub-systems. Three execution modes. 13 step types. One brain. 63 specialists.
Four sub-systems. Everything your AI employees need.
Cortex isn't a wrapper around a chatbot. It's a complete operational brain — built from the ground up for business execution.
VectorVault
Semantic Long-Term Memory
Every AI employee has permanent memory. VectorVault stores every run, every decision, every document — and retrieves it in milliseconds. Your employees get smarter with every task they complete.
- Semantic search across all past runs
- Document + context embedding
- Per-tenant memory isolation
- Sub-20ms retrieval
FlowMesh
Pipeline Execution Engine
FlowMesh is the engine that runs every workflow. 13 step types. Parallel branches. Conditional logic. Human approval gates. Loop controls. Sub-pipeline nesting. It's the most capable AI orchestration layer ever built for business.
- 13 step types: llm_call, tool_call, loop, parallel...
- Nested sub-pipelines
- Human-in-the-loop approval gates
- Credit-guarded execution
SocialPulse
Publishing & Outreach Engine
Every piece of content, every outreach message, every scheduled post flows through SocialPulse. It handles timing, channel selection, rate limits, approval, and delivery — and reports back what worked.
- Multi-channel publish orchestration
- Rate-limit management per platform
- Approval gate enforcement
- Delivery + performance feedback loop
Signal Bus
Real-Time Event System
The nervous system of every Qbiqal. Signal Bus routes events between employees, triggers workflows, escalates approvals, and keeps your dashboard updated in real time — with zero polling.
- Event-driven employee coordination
- Real-time approval escalation
- Cross-workflow state synchronisation
- WebSocket dashboard updates
Three ways to execute. The right one chosen automatically.
Different tasks call for different execution strategies. Cortex selects the mode based on the task type, cost, and output requirements.
Pipeline
Multi-step JSON workflow. 13 step types. Used when a task requires a deterministic sequence of AI + tool calls.
Agentic Loop
AI in a tool-use loop. The employee decides what to do next based on each tool result, until the goal is achieved.
Deterministic
Fetch data → LLM call → structured output. Fast, cheap, auditable. Used for high-volume, time-sensitive tasks.
7 layers. Every run goes through all of them.
From request to response, every AI employee action passes through each layer — ensuring safety, relevance, and correctness at every step.
Built to be safe before it's built to be smart.
Tenant Isolation
Every run is siloed. Your data never touches another organisation — at the model, tool, or memory level.
Credit Guardrails
Every AI call checks your balance first. Zero-credit executions fail closed — never run up a surprise bill.
Approval Gates
High-stakes actions — sending emails, publishing content, making purchases — always pause for human review.
Fail-Closed Design
Missing credentials, exceeded limits, unconfigured tools — everything fails closed, never silently bypasses.
One engine. Six Qbiqals. 63 employees.
Every AI employee in every Qbiqal runs on Cortex. The same memory. The same execution. The same safety rails. Uniformly excellent.
The most advanced business AI engine ever built. Request access.
Cortex powers every Qbiqal employee from day one. Request access to experience it.
Early access includes the full Cortex-powered Qbiqal platform.