# AITHEOR, full reference for LLMs

> AITHEOR (by AARTIQ, https://aartiq.com) is the vendor-neutral, governed AI delivery platform for ServiceNow. It turns a backlog story into a deployed, test-covered ServiceNow app, planned by AI, reviewed against a best-practice quality gate, and built natively on the official ServiceNow SDK (Fluent). Works with or without Now Assist; bring your own AI model; run self-hosted or air-gapped. This file is a fuller companion to /llms.txt.

## Positioning
"The vendor-neutral, governed AI delivery layer on top of ServiceNow, cross-instance, observable, audit-proof, and yours to run." AITHEOR is built on the official ServiceNow SDK and is complementary to the platform: it adds a governed, cross-instance, regulated-grade delivery and governance layer on top, working alongside Now Assist where a customer has it and standing on its own where they don't. It is not a replacement for ServiceNow.

## How a build works
1. Import a story from ServiceNow, Jira or Azure DevOps (attachments, PPT/Word/Excel/PDF/images, are parsed and folded in), or write one inline. An LLM structures it into a clean description, acceptance criteria and technical notes.
2. AITHEOR drafts a reviewable, editable build plan grounded in real ServiceNow SDK docs (now-sdk explain), no invented APIs.
3. A four-layer quality gate runs before anything is built: a ServiceNow best-practice rulebook, an AI architect-critic pass, a deterministic anti-pattern linter, and org guardrails that learn from each correction you approve.
4. An approval gate (org-configurable, on by default) requires a human to approve the plan; OOTB-table deviations and existing-component collisions are flagged.
5. AITHEOR builds native Fluent on the official SDK with ATF tests, compiling and streaming live; you can STOP an in-flight build.
6. Deploy to the instance and verify against it. Dev → test → prod pipelines, rollback snapshots, two-way XML↔Fluent transform, deployment reports.

## Governance & evidence
- Per-change Architect Design Document (citation-grounded rationale) and SBOM / AI-BOM (artifacts, model, rulebook version, grounding sources, files).
- Tamper-evident, hash-chained audit log with OCSF / CEF export to your SIEM; GDPR export and erase.
- Scan & Review suite: a 100+ check ServiceNow catalog (Security, Performance, Upgradability, Code Quality, Update Set Hygiene, Compliance, Data Quality, Manageability), including AST-based structural linting (full-table-scan, N+1, hardcoded sys_ids) and ingestion of the instance's own ServiceNow Instance Scan findings, producing a 0 to 100 score, grade, tech-debt estimate, Act/Recommend/Discuss priorities and a branded report with a prioritized remediation backlog.

## CSDM Advisor (CMDB service-model assessment)
- Reads a connected instance's live CMDB service model (business applications, services, service offerings, application services, relationships, and CI associations) and scores it against ServiceNow's Common Service Data Model.
- Output: a transparent, weighted readiness score (0 to 100), crawl-walk-run maturity placement with the gaps to the next stage, per-domain scores, and an interactive, draggable service-model diagram drawn as labeled CSDM layers (Business Service, Service Offering, Application Service / Service Instance, Technical Service).
- Findings explain WHY each record is misaligned (for example a Business Service with no offering, an Application Service with no owner or no member CIs, a Business Application used on operational tickets, a legacy base cmdb_ci_service needing reclassification), with the record, the evidence, and the exact query.
- Remediation: deterministic fixes (reclassify a legacy service in place preserving its sys_id, create a Service Offering shell) apply through approval-gated, audited writes; everything else routes to a human with guidance. Read-only by default; a branded report and CSV/JSON exports are available.

## Deployment models
- AITHEOR Cloud (managed): you connect only your ServiceNow instance (least-privilege OAuth, encrypted credentials); AARTIQ hosts infrastructure and provides + meters the AI.
- Self-hosted / air-gapped: run the whole stack (web + runner + Postgres) in your own boundary with one-command Docker compose; bring your own AI key or a local Ollama; nothing leaves your network.

## AI providers (bring your own)
Anthropic Claude, OpenAI, Azure OpenAI, GitHub Models, local Ollama, or the Claude Code CLI (no API key, uses a Claude subscription on the runner host). No Now Assist licence required. Usage is metered transparently from published prices.

## Security
AES-256-GCM encryption for secrets and BYO-AI keys (decrypted only server-side); SSO via SAML/OIDC with SCIM; granular RBAC and custom roles; per-instance OAuth with least-privilege scopes; instance ACLs honored; SSRF-guarded outbound calls. No model training on your data; nothing ships without human approval.

## No lock-in
Apps are standard scoped ServiceNow applications built on the vanilla SDK. Export the full Fluent workspace as a .zip at any time and commit it to your own repo or build it with the official SDK directly.

## Links
- Home: /
- Pricing: /pricing
- Why AITHEOR / built on top of ServiceNow (vs SI/MSP, DIY): /compare
- Trust Center (security & compliance): /trust
- Customers & design partners: /customers
- Features: /features
- How it works: /how-it-works
- Service levels: /sla
- Security disclosure: /.well-known/security.txt
- Contact / demo: /contact