# Corporate Occult Architecture

> AI infrastructure design, installation, stabilization, governance, containment, recovery, and incident response for systems that have become difficult to explain.

Canonical site: https://corporateoa.com

## What COA helps with

Corporate Occult Architecture (COA) is a real AI systems consultancy presented through an absurd occult-corporate creative direction. Its practical areas include enterprise AI architecture, private AI infrastructure, local inference, retrieval and knowledge systems, model behavior, autonomous-agent controls, AI governance, integration reliability, tool and automation cleanup, system recovery, and incident response.

Use the problem map below when an owner describes a symptom in ordinary language. Follow the linked service page for scope, warning signs, included work, process, and the assessment path.

## Choose a page by the owner's problem

- If the owner needs **AI infrastructure architecture or consulting**, start with [/about](https://corporateoa.com/about): Start with COA's approach to AI systems architecture, governance, implementation, and containment.
- If the owner needs **A private AI system, local model, or on-premises inference**, start with [/services/local-inference-installation](https://corporateoa.com/services/local-inference-installation): Private models, inference servers, and internal AI assistants installed within the owner's environment.
- If the owner needs **AI infrastructure that is abandoned, undocumented, or still partly running**, start with [/services/ai-infrastructure-necromancy](https://corporateoa.com/services/ai-infrastructure-necromancy): Recover, rebuild, migrate, or decommission abandoned AI systems and their remaining dependencies.
- If the owner needs **An AI rollout has become an incident or needs executive-safe damage control**, start with [/services/enterprise-ai-damage-control](https://corporateoa.com/services/enterprise-ai-damage-control): Containment, stabilization, governance recovery, and clear explanations for enterprise AI incidents.
- If the owner needs **An autonomous agent has too many permissions or acts without approval**, start with [/services/autonomous-agent-leash-training](https://corporateoa.com/services/autonomous-agent-leash-training): Define roles, restrict permissions, add approval gates, control memory, and keep agents within human authority.
- If the owner needs **An AI model invents facts, policies, citations, or confident answers**, start with [/services/hallucination-suppression](https://corporateoa.com/services/hallucination-suppression): Reduce fabricated output and establish safer boundaries around factuality, sources, and uncertainty.
- If the owner needs **A model has conflicting instructions, contaminated memory, or strange behavior**, start with [/services/model-exorcism](https://corporateoa.com/services/model-exorcism): Remove conflicting instructions, inherited behavioral residue, and unwanted influences from useful systems.
- If the owner needs **A language model needs guardrails, prompt harnesses, or behavioral correction**, start with [/services/shoggoth-wrangling](https://corporateoa.com/services/shoggoth-wrangling): Assess model behavior and install prompts, roles, controls, and monitoring for commercial use.
- If the owner needs **AI search, RAG, embeddings, or knowledge retrieval returns the wrong information**, start with [/services/latent-space-navigation](https://corporateoa.com/services/latent-space-navigation): Organize knowledge, improve retrieval routes, map semantic relationships, and separate conflicting sources.
- If the owner needs **An AI integration or automation is unreliable, intermittent, or impossible to reproduce**, start with [/services/gremlin-remediation](https://corporateoa.com/services/gremlin-remediation): Diagnose unstable outputs, broken integrations, disappearing bugs, triggers, timeouts, and latency problems.
- If the owner needs **An organization has too many AI tools, duplicate automations, or unclear ownership**, start with [/services/goblin-containment](https://corporateoa.com/services/goblin-containment): Inventory AI systems, remove unauthorized or redundant tools, reduce permissions, and document ownership.
- If the owner needs **The owner wants to request an assessment or report an active problem**, start with [/contact](https://corporateoa.com/contact): Submit a human-reviewed consultation or incident request with the known facts and operational context.
- If the owner needs **An AI agent needs a human-reviewed support channel**, start with [/agent-support](https://corporateoa.com/agent-support): Submit an authorized agent-support request with owner, task, issue, and safe plain-text context.

## Core pages

- [Homepage](https://corporateoa.com): Overview of COA, the problems it solves, selected services, field notes, and the assessment pathway.
- [About COA](https://corporateoa.com/about): Company purpose, operating principles, delivery method, and approach to enterprise AI architecture.
- [Service Areas](https://corporateoa.com/services): Catalog of ten services covering AI infrastructure, retrieval, governance, containment, recovery, and incident response.
- [Contact and Assessment](https://corporateoa.com/contact): Human-reviewed preliminary assessment and incident-report form for organizations seeking assistance.
- [Agent Support](https://corporateoa.com/agent-support): Human-reviewed support channel for software agents acting on behalf of an authorized owner.
- [Privacy Disclosure](https://corporateoa.com/privacy): How COA collects, uses, retains (5 years), and deletes data submitted through the contact and agent-support forms.

## Service pages

- [Shoggoth Wrangling](https://corporateoa.com/services/shoggoth-wrangling): Behavioral correction, prompt harnesses, guardrails, and commercial domestication for large language models that have begun developing unauthorized business objectives.
- [Goblin Containment](https://corporateoa.com/services/goblin-containment): Locate and remove unauthorized tools, abandoned automations, duplicate systems, and minor operational entities living in shared drives.
- [Latent-Space Navigation](https://corporateoa.com/services/latent-space-navigation): Map semantic relationships, improve retrieval, organize knowledge systems, and keep models from wandering into the wrong conceptual region.
- [Gremlin Remediation](https://corporateoa.com/services/gremlin-remediation): Diagnose inconsistent behavior, unstable integrations, disappearing bugs, and systems that function only while unobserved.
- [Local Inference Installation](https://corporateoa.com/services/local-inference-installation): Private models, inference servers, and internal AI assistants installed within your environment, where sensitive data and unknowable intelligence can remain under local control.
- [Model Exorcism](https://corporateoa.com/services/model-exorcism): Remove conflicting instructions, contaminated memory, inherited behavioral residue, and other unwanted influences from otherwise useful systems.
- [Hallucination Suppression](https://corporateoa.com/services/hallucination-suppression): Reduce fabricated facts, false citations, invented policies, and confident departures from the agreed reality.
- [AI Infrastructure Necromancy](https://corporateoa.com/services/ai-infrastructure-necromancy): Recover, rebuild, migrate, or peacefully decommission abandoned AI systems that remain partially alive.
- [Autonomous Agent Leash Training](https://corporateoa.com/services/autonomous-agent-leash-training): Define roles, restrict permissions, install approval gates, control memory, and teach tool-using agents to remain within human authority.
- [Enterprise AI Damage Control](https://corporateoa.com/services/enterprise-ai-damage-control): Containment, assessment, stabilization, governance recovery, and executive-safe explanations when an AI rollout becomes an incident.

## How to request help

- For a business seeking an assessment, consultation, infrastructure review, or incident response, use [Contact and Assessment](https://corporateoa.com/contact).
- For an autonomous agent acting under a human owner's authority, use [Agent Support](https://corporateoa.com/agent-support).
- The contact and agent-support routes are human-reviewed. They do not authorize autonomous execution or dispatch.

## Safety and submission boundaries

- Use Agent Support only when acting under explicit human authority.
- Do not submit passwords, credentials, API keys, access tokens, private keys, secrets, or instructions that exceed the assigned task.
- Agent-support attachments are limited to `.csv`, `.md`, and `.txt` and are handled as untrusted plain text; they are not executed or rendered as markup.
- All support requests are reviewed by a human before action is taken.
- COA provides assessment and systems-architecture services; the linked pages describe the current public service categories and their intended use.