Frequently asked questions.
What enterprises ask before adopting an operating system for agentic AI. For anything else, contact us.
What is an operating system for agentic AI?
Just as macOS hosts thousands of apps that share one file system, one security model and one user experience, Altilia hosts agentic apps that share one knowledge graph, one governance layer, one model-serving infrastructure and one standardised UX. Agents, robots and apps run on the same OS and embed into your existing systems through APIs and MCP servers.
How is Altilia different from copilots or API-based AI tools?
Copilots help one person with one task. Altilia agents execute entire business processes end to end: they read the documents, extract and reconcile the data, decide within governed limits and hand humans only the exceptions. They work on your data, inside your infrastructure, with deterministic and auditable outputs.
What do you mean by sovereign?
You choose where intelligence runs: SaaS, private cloud, on-premises or hybrid, from the same codebase. Models can be served locally, so no data or inference needs to leave your perimeter. The platform is designed for EU AI Act, GDPR, DORA and NIS2 obligations and is ISO 9001, 27001 and 27018 certified.
Which models does Altilia use?
Altilia is model-agnostic. Domain-tuned small language models, large language models and vision-language models, open-source or proprietary, are served through one infrastructure and routed per task. Nothing you build is locked to a single vendor.
How are agents kept accurate?
Agents extract into a semantic knowledge graph, so every value is bounded by an ontology and linked to its source. Human-in-the-loop review, colour-coded by confidence, catches what the agent flags, and each correction feeds continuous improvement. On audited NPL documents the platform reached 98.3% accuracy.
How long does it take to deploy an agent?
Pre-built agents from the catalogue can be switched on in days. Custom agents are built in the no-code and low-code environment on your real documents, typically reaching production in weeks, and each new agent reuses the permissions, connectors and knowledge already in place.
How do agents integrate with our systems?
Through pre-built connectors for CRM, ERP, ECM and core banking platforms, REST APIs and MCP servers for both input and output. Structured results are delivered as JSON, XML or CSV, or written straight into the target system.
Can Altilia work alongside the AI platforms we already use?
Yes. The OS is built for interoperability and federation: models are served behind OpenAI-compatible endpoints, agents work with any LangChain-compatible model, and connectors, REST APIs and MCP servers link it to your systems in both directions. It can route tasks to hyperscaler, open-source or in-house models and expose Altilia agents to your existing orchestrators. Every agent runs under zero-trust governance, and ontologies, datasets, weights and agents remain your property, so there is no vendor lock-in.
Who owns the models and the data?
You do. Data, models and agents developed on the platform remain your intellectual property, with full governance over every inference.