★ status: autonomous operations active / agent v0.0.1 deployed across sector / twin sync: 99.97% / drift reconciliation: continuous / sn-0047-gcx · batch 004-a / classification: enterprise / field-deployed /
operational category · simulation-driven autonomous itops build [01] · v 0.0.1 origin · london, uk 2026-04-26 14:32:07 utc

The self-driving
layer for IT operations.

General Cognitions is building simulation-driven autonomous ITOps: digital twins and agents that hold a live model of the estate, rehearse every change against it, and execute through the tools your team already runs. No more keyboard handed back to a human at the moment of action.

★ deployed alongside the systems already running in your estate / ConnectWise · Datto · Kaseya · NinjaOne · Splunk · CrowdStrike · Veeam · Microsoft 365 /
01 / sector report
[ field analysis ]

The labour arithmetic is broken.

IT estates have outgrown the humans operating them. Headcount cannot scale with estate complexity, and the tools meant to help (monitoring, AIOps, traditional automation) still require a human to act on the alert.

Every CIO survey in the last three years places “automation of operations” in the top three priorities. The Gartner AIOps market is roughly $2B and growing ~15% annually; the broader IT operations software market is north of $30B. The spend is there. The products under-deliver because they stop short of execution.

The missing primitive is a live, high-fidelity model of the estate that autonomous agents can reason against, rehearse in, and act through, safely.

it operations software
$30B+

Global annual spend on tooling that still hands the keyboard back to a human.

aiops market growth
~15%

Compounding annually, a $2B segment chasing the autonomous-execution gap.

cio priority ranking
Top 3

Three years running, “automation of operations” sits in the top three priorities.

02 / state of the stack
[ current architecture ]

Reactive. Fragmented.
Human‑dependent.

Modern IT infrastructure is too complex to operate manually. The tools meant to help (observability, ticketing, scripting) are reactive by design and depend on a human to close the loop.

Today's operational stack is three layers in series, with a person standing in the gaps. Each handoff imposes latency, judgement load, and risk. Drift returns. The chain runs again.

[ today's operational chain / fig 02 ] 3 layers · 2 human gaps · loop never closes
┌──────────────────────────────────┐     ┌──────────────────────────────────┐     ┌──────────────────────────────────┐
│OBSERVABILITY · METRICS / SIEM    │     │HUMANS · NOC · CONTEXT · ROUTE    │     │SCRIPTS/TOOLS · RMM · RUNBOOK     │
│INGEST · NORMALIZE · ALERT STORM──▶TRIAGE · CHAT · TRIBAL KNOWLEDGE──▶EXECUTE PLAYBOOK · PARTIAL CONTROL│
│"SOMETHING IS WRONG" · NO OWNER   │     │"FIGURE IT OUT" · GUESS / ROUTE   │     │"TRY TO FIX IT" · HOPE / PRAY     │
│SIGNALS RICH · CAUSALITY POOR     │     │QUEUES DEEP · MODEL ABSENT        │     │PATCH SHIPS · DRIFT RETURNS       │
└──────────────────────────────────┘     └──────────────────────────────────┘     └──────────────────────────────────┘
         ▲                                                                                                           │
         │                                                                                                           ▼
         └─────────────────────────────────DRIFT RETURNS · HUMAN GAPS · LOOP REOPENS─────────────────────────────────┘
[ what's missing ]

a fourth layer · the one that closes the loop

A system that understands the environment and can safely act on it.

Not a faster ticket queue. Not a smarter dashboard. A simulation-driven autonomous layer that holds a live model of the estate, rehearses every change against it, and executes through the tools your team already runs.

03 / protocol
[ multi-layer architecture ]

Four layers.
One autonomous loop.

The platform is built as a stack. Each layer feeds the one above it. Telemetry becomes a model. The model becomes a rehearsal surface. The rehearsal becomes a safe production change. The change becomes new telemetry.

L1 / 01

Data Ingestion

Continuous capture from PSA, RMM, SIEM, cloud control planes, identity, backup, network, and endpoint telemetry. Normalised into a single estate graph.

ingest · stream
112 connectors
L2 / 02

Digital Twin

A live, high-fidelity model of the estate covering assets, dependencies, identities, policies, configurations, and the desired state the agent reconciles against.

twin · graph
fidelity 99.97%
L3 / 03

Execution Environment

A sandboxed plane where agents rehearse changes against the twin, simulate blast radius, and validate outcomes before any action touches production.

rehearse · simulate
blast-radius bounded
L4 / 04

Autonomous Agents

Specialised agents reason against the twin, plan, rehearse, then dispatch changes through the same RMM, automation, and scripting tools your team already operates.

execute · reconcile
supervised → autonomous
>>>>>> data-flow · ingestion → twin → rehearsal → execution → reconciliation protocol-id : gc-pr-04 / build [01]
04 / capabilities
[ operational primitives ]

Autonomous change.
Continuous reconciliation.
Active remediation.

★ 01 / change-execution

Autonomous Change Execution

Plan, rehearse, and dispatch changes across complex IT estates without handing the keyboard back to a human. Every action is rehearsed against the twin before it runs.

★ 02 / drift-reconcile

Desired-State Drift Reconciliation

A continuous loop. The agent compares observed state to desired state across the estate and reconciles the gap across configuration, policy, identity, patch level, backup posture.

★ 03 / remediation

Incident Remediation

When a SIEM, RMM, or endpoint signal fires, the agent triages against the twin, determines blast radius, executes the safe path, and writes the action back into the estate record.

05 / integrations
[ connector matrix ]

Compatible with the systems already running in your estate.

The agent acts through the tools your team already pays for. We do not replace your stack; we operate it. Below is a non-exhaustive matrix of supported connector classes and representative vendors.

class surface area representative vendors status
Professional Services Automation (PSA) service desk & ticketing ConnectWise · Autotask · HaloPSA · SuperOps
Remote Monitoring & Management (RMM) endpoint fleet & agents NinjaOne · Datto RMM · Kaseya VSA · N-able · Atera
Security Information & Event Management (SIEM) security telemetry Splunk · Microsoft Sentinel · Elastic · Sumo Logic
IT Automation & Orchestration orchestration & scripts Ansible · PowerShell · Bash · Rewst · Tines
Cloud & Hybrid Data Backup data protection Veeam · Acronis · Druva · Rubrik
Business Continuity & Disaster Recovery (BCDR) disaster recovery Datto BCDR · Axcient · Cohesity · Zerto
Email Security & Secure Email Gateway email gateway Proofpoint · Mimecast · Abnormal · IRONSCALES
Unified Communications & Collaboration collab & scheduling Microsoft 365 · Google Workspace · Slack · Teams
Identity & Access Management (IAM) directory & access Okta · Entra ID · JumpCloud · Duo
connector classes · 9 · vendor endpoints · 112 · open / unrestricted spec-id : gc-int-04 · last reconciled today
general cognitions · sn-0047-gcx · build [01] classification: open / unrestricted
06 / thesis
[ desired-state model ]

A live model the agent can reason against, rehearse in, and act through.

[ field manual ]

excerpt · chapter 04 · pages 12–14

Monitoring tells you something is wrong. AIOps clusters the noise. Traditional automation runs the script you wrote last quarter. None of them close the loop without a human in the chair.

General Cognitions ships the missing primitive: a desired-state model of the IT estate that an agent continuously reconciles against reality. Every change is rehearsed against the twin before it runs. Every reconciliation is written back. Every action is auditable. The keyboard does not get handed back.

This is the first autonomous change-execution platform: a self-driving layer for IT operations that executes changes, reconciles drift, and remediates incidents without human intervention.

[ closed-loop autonomy / fig 04 ] no human in the chair · agent reconciles continuously
   ┌──────────────────────────┐                                                           ┌──────────────────────────┐
   │LIVE ESTATE · RUNTIME     │ ───────────────── DRIFT · DIFF · RECON ──────────────────▶│DIGITAL TWIN · GRAPH      │
   └────────────┴─────────────┘                                                           └────────────┬─────────────┘
                ▲                                                                                                    │
                │ EXECUTE · REHEARSED · AUDITED                                              REASON · PLAN · CONSENT │
                │ NO SHADOW RUN · TOOL BRIDGE ONLY                                                                   ▼
   ┌────────────┬─────────────┐                                                           ┌────────────┴─────────────┐
   │CHANGE / ACTION · TOOLS   │◀──────────── REHEARSE · BLAST · CONSENT GATE ─────────────│AUTONOMOUS AGENT · CORE   │
   └──────────────────────────┘                                                           └──────────────────────────┘
[ engage operations ]

Hand the keyboard to the agent.

We are deploying the platform with a small set of partner estates running complex, multi-tenant IT operations. If that describes you, we want to hear about your environment and the changes you are tired of running by hand.