FAQ
Why predictable AI runtime control?
The questions investors and customers are most likely to ask.
- Why does AI execution need to be predictable?
Large AI models are probabilistic by nature. Given the same input, they can produce different outputs, interpret context differently, or propose different actions.
That is acceptable when AI is generating text, images or recommendations. It becomes a problem when AI is executing actions that affect real systems, data, money or physical assets.
Relayn introduces an operating system that makes AI in the physical world more predictable: models propose, approved rules govern what actually runs.
Models reason probabilistically. Rules decide. Systems execute.
The objective is not to remove AI's intelligence. It is to make the consequences of AI decisions controlled, predictable, repeatable and auditable.
- What does “deterministic” actually mean in Relayn?
In Relayn, “deterministic” describes the rule engine’s execution path under your approved rules — not a claim that the whole system never varies. Once inputs, policies and conditions are defined, that path is controlled and repeatable.
The AI can propose an action, but it cannot arbitrarily change the rules governing execution.
For example:
- AI: “Reduce the cooling system by 15%.”
- Relayn determines whether that action is permitted, under which conditions, within what limits, and what happens next.
The same defined conditions lead to the same controlled execution.
This creates a fundamental separation:
Probabilistic reasoning → rule-led decision and execution → real-world action.
- What is the proof behind Relayn OS cost efficiency?
Two measurements from our own use. In July 2026, 71% of 4,026 production requests were handled by rules, with no language-model call. Comparing billing for February to April 2026 with our earlier model-first version, AI inference spend fell by 95%. See the evidence
- How is Relayn different from traditional workflow or business-rule engines?
Workflow engines are generally designed around predefined processes.
For example: If A happens → execute B → then execute C.
Relayn addresses a different problem: AI-generated decisions entering an operational environment where the decision itself may be probabilistic.
The challenge is therefore not simply automating a known workflow. It is controlling the transition:
AI reasoning → validated decision → governed execution.
Relayn combines the flexibility of AI reasoning with predictable
- How is Relayn different from traditional cybersecurity, policy or governance tools?
Security and governance tools are usually separate products that decide who can access what, or whether an action complies with a policy. In Relayn, the rules are part of the operating system itself. The agents run inside it, on your systems, and you control them directly. There is nothing extra to integrate.
- What is the fundamental difference between Relayn and today's AI infrastructure?
Most AI systems run in a line: prompt, model answer, tool. Relayn OS runs in a loop. It observes your systems, builds context, checks each proposed action against your rules, and feeds outcomes back. The model is one part of that loop.
- The simplest way to explain Relayn
Relayn is an operating system that lets AI act in the physical world, under your rules.
- Can Relayn run on our own infrastructure?
Yes. Hosted is Relayn-operated cloud; managed and enterprise run on your servers, private cloud or on your premises. Data and models stay under your policies. Compare deployment options
- Do we have to replace the AI tools or business systems we already use?
No. People keep the assistants and channels they prefer; Relayn connects to your systems and knowledge once, and every agent reuses the same permissions. Legacy systems without an API can still be operated through their own screens. See Work and Connect.
- Do you have security certifications?
Not yet. We are a young company. We support your own security review with an architecture description, answers to your questionnaire and a data processing agreement. Security
- Which models do you use? Can we use our own?
Relayn OS is model-agnostic. It runs open models on your own hardware and can use cloud models where you allow them. You can bring your own models and keys. Connect