Three things we got right before anything else.
Encryption at rest & in transit
All tenant data — documents, metadata, vector embeddings, audit records — encrypted with AES‑256 under customer-managed keys in your KMS. TLS 1.3 end-to-end. Qualitum never holds your encryption keys.
Identity & authentication
SAML 2.0, OIDC, OAuth 2.0, LDAP. Native connectors for Microsoft Entra ID, Okta, Ping, and Active Directory with SCIM provisioning. MFA enforced at platform level. No parallel user directory. No shadow accounts.
Five-tier role-based access
Platform Admins → Process Owners → Knowledge Owners → Validators → Agent Users. Every permission enforced at document level. RAG retrieval respects RBAC at query time — agents cannot surface what the user cannot see.
Inside your tenant. Your cloud. Your region. Your keys.
Your data never leaves your perimeter.
Qualitum is a single-tenant deployment model. Each customer runs in a dedicated environment on their chosen cloud, in their chosen region, under their IAM. There is no shared data store, no shared model state, and no shared inference pool between customers — ever.
- CloudAWS · Azure · GCP · on-prem. Bring your own. Air-gap supported for classified workloads.
- RegionAny. EU (Frankfurt, Ireland, Paris), US, UK, UAE, Singapore, Sydney — pinned to your residency requirement.
- NetworkYour VPC. Private subnets, VPC endpoints, PrivateLink. No public ingress required for the agent runtime.
- StorageYour buckets, your database. Object storage in your S3/Blob/GCS. Postgres in your RDS/Cloud SQL. Your snapshots, your backups, your retention.
- KeysYou hold them. Customer-managed KMS keys (AWS KMS, Azure Key Vault, GCP KMS). Revoke anytime and the data becomes unreadable — to us, to you, to anyone.
- EgressZero. Inference happens inside your VPC. No prompt content, no completions, no embeddings, no telemetry leaves your network.
Copilot is a chat surface. Qualitum is operational agents.
Different primitive. Different problem. Different outcome.
Microsoft Copilot, ChatGPT Enterprise, and Gemini Enterprise are general-purpose assistants: a chat box over your documents. Qualitum is not a chat box. It is a set of domain-specific, validated operational agents that execute regulated workflows end-to-end. Here is the practical difference.
| Dimension | Microsoft Copilot · ChatGPT Enterprise | Qualitum |
|---|---|---|
| Primitive | Chat box over your M365 / corporate data. User asks, model responds. | Validated operational agents that author, execute, and defend a specific regulated workflow. No human prompt engineering required. |
| Deployment | Multi-tenant SaaS. Your data in Microsoft's / OpenAI's cloud, under their keys, subject to their roadmap. | Single-tenant. Your cloud, your region, your keys, your IAM. Air-gap capable. |
| Inference location | Shared inference pool. Prompts processed in the vendor's shared environment. | Inference runs inside your VPC. Zero egress. Every token stays in your perimeter. |
| Model | Vendor-chosen. OpenAI GPT for Copilot, OpenAI for ChatGPT Enterprise. You cannot swap it. | LLM-agnostic. Azure OpenAI, Anthropic Claude, Gemini, Mistral, LLaMA, or your in-house fine-tune. Swap per workflow. |
| Training on your data | Contractual opt-out. Data handling governed by the vendor's terms, revisable by the vendor. | Architecturally impossible. There is no pipeline out. Model weights sit inside your tenant. |
| Retrieval security | Graph permissions honoured, but enforced by the vendor's RAG. You trust their implementation. | 5-tier RBAC enforced at retrieval, inside your tenant. Every chunk gated by the user's identity before it reaches the model. |
| Audit trail | Usage logs. Limited traceability of which document informed which answer. | Tamper-evident ledger. Every retrieval, every token, every edit, every approval — immutable, signed, exportable to regulators. |
| Regulatory posture | Generic enterprise compliance. Not designed for GxP, Annex 11, 21 CFR Part 11, or EU procurement directives. | Purpose-built for GxP, Annex 11, 21 CFR Part 11, GAMP 5, EU AI Act, Directive 2014/24/EU. Validated agent outputs. |
| What you buy | Per-user monthly subscription. Price scales with headcount. Vendor owns the software. | Perpetual licence on your deployment. You own the software, the data, the model weights, and the modifications. |
The two are not competitors — they solve different problems. Most Qualitum customers continue to run Copilot for general productivity. Qualitum handles the regulated work Copilot is not architected for.
The architecture behind Qualitum.
Private deployment · LLM agnostic · data sovereign · enterprise-grade. Every layer engineered for the environments Copilot-class tools were not built for.
A private agentic runtime, not an LLM wrapper.
Qualitum is the agentic layer. It combines domain knowledge, deterministic workflow, and context isolation into auditable agents that execute regulated work. The LLM is a component we swap. The agent behaviour — what gets drafted, how it is checked, what the audit trail looks like — is ours.
Domain knowledge
Agents pre-trained on industry-specific process patterns, regulatory requirements, and domain terminology — GxP, procurement law, EU AI Act.
Deterministic workflow
Process rules layered above the LLM ensure consistent, auditable outputs — regardless of which model is running underneath.
Context isolation
Each agent operates in a fully isolated context. No data bleed between agent types, no shared conversation state across tenants.
LLM agnostic by design.
Qualitum is not tied to any single model. Bring your preferred LLM — Azure OpenAI, Anthropic Claude, Google Gemini, a self-hosted Mistral or LLaMA, or a custom fine-tune — or use the default private model runtime that ships with the platform. Route different workflows to different models.
Azure OpenAI
EU data boundary, Microsoft Entra ID integration, private endpoint.
Anthropic Claude
Claude via AWS Bedrock or direct enterprise API with zero data retention.
Google Gemini
Gemini via Vertex AI in your GCP project, no cross-project data flow.
Self-hosted
LLaMA 3, Mistral, Qwen, or your in-house fine-tune on your GPU fleet.
Deployed in your infrastructure.
Fully automated CI/CD pipeline. From contract to live production agents in under 14 days for standard connectors. All environment provisioning defined as code — repeatable, version-controlled, auditable.
Isolated tenants
Each client runs on a fully isolated tenant. No shared infrastructure, no shared data stores, no cross-client exposure of any kind.
Cloud agnostic
Deploy on AWS, Microsoft Azure, Google Cloud, or on-premises. The pipeline is cloud-neutral and infrastructure-agnostic.
Air-gap capable
For classified or high-security environments, Qualitum can be deployed in a fully air-gapped configuration with no external network dependency.
Data sovereignty & privacy.
Data sovereignty is architectural — not a contractual promise. Qualitum is designed from the ground up so that client data never leaves client infrastructure. There is no telemetry pipeline, no training on client data, and no shared model state between tenants at any point.
AES‑256 encryption
All data encrypted at rest with AES‑256. Customer-managed keys — you hold the key, Qualitum cannot access your data. Revoke at will.
Perpetual licence
You own the software licence permanently. No SaaS subscription to be cancelled, no data held hostage by a vendor.
Tamper-evident audit log
Every agent interaction logged with immutable timestamps. Full audit traceability for regulatory inspections and internal governance.
Authentication & access control.
Enterprise-grade identity and access management out of the box. Qualitum integrates with your existing identity provider — no parallel user directory, no shadow credential management, no separate login portal. If someone loses access in your IdP, they lose access to Qualitum the same second.
SSO / LDAP
Active Directory, Entra ID, Okta, Ping, any LDAP-compatible IdP via SAML 2.0 or OIDC.
5-tier RBAC
Granular roles enforced at the retrieval layer — not just the UI. Governs which agents, data sources, and outputs each role can access.
MFA enforced
Multi-factor authentication enforced at platform level. Configurable per-tenant policy, per-role requirements.
SCIM provisioning
Automated lifecycle management. Accounts provision and deprovision automatically when employees join, move, or leave.
Connected to the systems of record.
Qualitum is the agentic layer above your stack, not a replacement. Agents read, write, and round-trip into your validated systems of record — eQMS, ERP, LIMS, procurement, clinical — without owning the source of truth. Bi-directional, auditable, idempotent.
eQMS
Veeva Vault Quality, MasterControl, Sparta TrackWise, Kneat Gx.
ERP
SAP S/4HANA, Oracle Fusion, Microsoft Dynamics 365.
Procurement
SAP Ariba, Coupa, Jaggaer, SAP Fieldglass.
LIMS / MES
LabWare, Siemens Opcenter, Werum PAS-X, Rockwell FactoryTalk.
You own the platform and the data.
The SaaS model was designed for software vendors, not for enterprises handling sensitive regulated knowledge. Qualitum issues a perpetual licence — you own the deployment, you own the data, and in most tiers you own the modifications. We win when your team grows into the platform, not when you get stuck on it.
-
Perpetual licence, no per-user fees
One licence covers unlimited users. No fees that compound as your headcount grows. No vendor renegotiation at renewal.
-
Modify and extend it yourself
Full source access under licence for enterprise tier. Your team can extend agents, customise workflows, and add integrations without involving us.
-
Your data belongs to you
Complete data portability. Export everything at any time — documents, audit trail, vector index, model fine-tune weights. No proprietary formats. No lock-in. If you ever leave, everything comes with you.
| Characteristic | Qualitum | Typical SaaS AI |
|---|---|---|
| Pricing model | Perpetual licence | Per-user / month |
| Data location | Your infrastructure | Vendor cloud |
| AI model on your data | Your model, private | Shared inference |
| Training on your data | Never — architectural | Often by default |
| Modifiable by you | Yes — full source | No |
| Air-gap deployment | Yes | No |
| Data portability | Complete | Limited / export fees |
Certified where it matters. Aligned where it is required.
Qualitum is built, operated, and audited against the regimes that govern the next decade of regulated work.
Everything your InfoSec team will ask.
No — and not because of a contractual clause. Because the pipeline does not exist. Inference runs inside your tenant; model weights never leave it; there is no upstream channel to Qualitum. Even if we wanted to, there is nowhere for the data to go.
The deployment is yours under the perpetual licence — it keeps running. You retain all data, all audit logs, all configurations, and all fine-tune weights. We can offer a 90-day transition package where we hand over the knowledge-transfer documentation and detach from operational support; after that your platform keeps operating without us.
You do. Keys live in your KMS (AWS KMS, Azure Key Vault, GCP KMS, or your on-prem HSM). We configure the platform to use them; we never have access to the key material. If you revoke the key, Qualitum can no longer read your data — and neither can anyone else.
No. Qualitum engineers have no standing access to your tenant. Break-glass support requires explicit, time-bound, customer-approved access through your IAM — every session logged in your audit trail, not ours. Most customers never grant it.
Only if you choose a hosted LLM and only under the contractual terms you sign with them. Most regulated customers run inference through Azure OpenAI in their own EU data boundary, Bedrock in their own AWS account, Vertex in their own GCP project, or a self-hosted open-weight model — in which case no external provider sees anything. Every option is configurable per workflow.
The architecture is the privacy control. No telemetry egress, no shared inference, no cross-tenant data flow, RBAC enforced at retrieval. Data minimisation is enforced at ingest (we index only what you scope). DPIA templates and sub-processor inventory ship with the platform.
Yes. Fully air-gapped deployments are supported for defence, public-sector classified environments, and pharma facilities with strict network isolation. A self-hosted open-weight model runs locally; updates are shipped via signed, offline artefacts; no outbound network dependency exists.
Qualitum deployments classified as high-risk AI systems ship with a conformity-assessment-ready technical file: risk management documentation, data governance records, logging, human oversight configuration, and transparency artefacts. We align to the 2026 enforcement timeline.
Yes. Enterprise-tier customers receive full source access under the perpetual licence, with a signed SCA (source-code access) agreement. Your security team can review, pentest, fork, and extend the platform. Most customers do not modify it; all of them can.
Sev-1 incidents: 15-minute acknowledgement, 4-hour mitigation target, post-incident RCA within 5 business days. Sev-2 and Sev-3 scaled accordingly. Every incident reported to you within the NIS2 / DORA disclosure windows applicable to your jurisdiction.
Bring your InfoSec team. Bring your hardest questions.
Book a working session with a Qualitum security engineer. We will walk through the architecture, the audit trail, and the exact answers to the questions your CISO is paid to ask.
Book a callOr email info@qualitum.ai