Every question we get before a working session, answered plainly.
No jargon left unexplained. This page carries the full FAQ plus a plain-English glossary of the AI terms used across the site, so you can walk into a working session already speaking the same language.
What is AutoServa?
AutoServa is VYROX's workflow automation service. We consult, design, deploy and implement agentic AI, fine-tuned on how your organisation actually decides, that works between your business and its workflow while your team keeps approval.
Does AutoServa replace our existing software?
No. AutoServa works your existing stack, including ERP, CRM, accounting, mail, document stores and the browser based portals nobody ever built an integration for. There is no migration, no rebuild and no new screen for your staff to learn.
How is this different from a chatbot or a generic AI assistant?
A generic assistant answers from what it read on the internet, in a plausible generic voice. AutoServa is fine-tuned on your own documents, decisions and vocabulary first, so its answer matches the one your team would recognise as the house answer, and it acts inside your systems rather than only replying in a chat window.
How is this different from RPA (robotic process automation) or a rules engine?
RPA repeats a fixed sequence of clicks and breaks the moment a case does not match the script. AutoServa reasons through a case the way a trained staff member would, applies your decision framework rather than a hard-coded path, and can handle a case it has not seen in exactly that shape before.
Which cloud AI providers does AutoServa support?
AutoServa is not locked to one vendor. We connect the operator to frontier models from OpenAI (GPT-5), Anthropic (Claude Sonnet 5 and Claude Opus 5), Google (Gemini 3), Moonshot AI (Kimi K2) and Zhipu AI (GLM-5.2), and move it to a new model the moment a better one ships.
Can AutoServa run entirely on our own local AI server?
Yes. For clients who cannot send data off-premise, we deploy open-weight models, including Qwen3, Meta Llama 4 and Google Gemma 3 for reasoning, on a server inside your own building or private cloud, fully air-gapped if your policy requires it.
Who stays in control of what AutoServa does?
Your team does. Workflow Guard checks approval chains, spending ceilings, access scope, segregation of duties, retention rules and regulatory obligations before any action executes, and nothing irreversible happens without an approval you defined.
What happens when the operator is not sure what to do?
It escalates. The decision framework defines explicitly when a case stops being routine, and the operator hands that case to a person with its reasoning and the relevant precedent attached, rather than guessing or stalling silently.
How long before we see the operator working live?
A decision audit runs two to three weeks. Design and deployment, tuning the operator and running it live beside your team, typically reaches a supervised live run within five to eight weeks of starting, depending on how much of the decision framework already exists in writing.
What does an engagement cost?
Pricing is quoted per engagement, based on the workflow's complexity and volume, not a seat licence. The decision audit is the cheapest way to get a firm number: you will know the shape of the cost before committing to design and deployment.
What do we keep if we stop after the decision audit?
The written decision framework for that workflow, in full: the inputs, thresholds, constraints, approvals and escalation rules we documented together. It is yours whether or not you continue to design and deployment.
Does our data get used to train models for other clients?
No. The operator is tuned specifically on your material and stays inside your infrastructure boundary, cloud connection or local server. It is never pooled with another client's data or used to improve a shared model.
What industries or company sizes does AutoServa fit?
AutoServa fits any organisation with a repetitive, reviewable, decision-heavy workflow, most commonly mid-size to large organisations in finance, legal, procurement, service operations and reporting. The determining factor is the workflow, not the industry label or headcount.
How does an engagement start?
With a decision audit: two to three weeks mapping how one workflow really decides, with the decision framework written down and yours to keep whether or not you continue.
How do we contact VYROX?
Message us on WhatsApp at +60 19 688 3338. There is no email enquiry channel or contact form; we reply on WhatsApp, usually the same working day.
The AI terms on this site, in plain English.
We use these words carefully and try to explain them the first time, but here they are in one place, defined the way we'd explain them out loud in a working session.
- Agentic AI
- AI that can plan a sequence of steps and take actions toward a goal, not just answer a single question. AutoServa is agentic: it reads a case, decides what to check next, and acts, rather than waiting for a person to prompt every step.
- Fine-tuning
- Further training a general AI model on a specific, narrower set of material, in this case your documents, decisions and vocabulary, so its answers reflect your organisation instead of a generic internet average.
- Decision framework
- A written structure describing the inputs a decision depends on, the thresholds that change the answer, the constraints that cannot be crossed, who approves what, and when a case must escalate to a person.
- Workflow Guard
- AutoServa's governance layer. It checks every action the operator would take against approval chains, spending ceilings, access scope, segregation of duties, retention rules and regulatory obligations before that action executes.
- Evidence trail
- The record attached to every decision the operator makes: what it read, which rule in the decision framework applied, what threshold was hit, what precedent it followed, and who approved the release.
- Precedent
- What was decided last time on a similar case, kept searchable so a new decision stays consistent with past ones instead of depending on whoever happens to be handling it.
- Frontier model
- The most capable AI models currently available from a given provider, for example GPT-5 from OpenAI or Claude Opus 5 from Anthropic. AutoServa is not tied to one; we connect the operator to whichever fits the workflow.
- Open-weight model
- An AI model whose underlying parameters are published and can be run on a client's own hardware, rather than only accessed through a provider's cloud API. This is what makes a fully local, air-gapped deployment possible.
- Local AI server
- A server inside a client's own building or private cloud running open-weight models, used when data cannot leave the premises. Same operator, same Workflow Guard, same evidence trail as a cloud deployment.
- Air-gapped
- A system with no network connection to the outside internet, the strictest form of local deployment, used where policy requires that no data can leave the premises under any circumstance.
- Human in the loop
- A design principle where a person must review or approve before an action is finalised. AutoServa keeps a human in the loop on anything irreversible; routine, in-policy steps can proceed without a fresh approval each time.
- Decision audit
- The first stage of an AutoServa engagement: two to three weeks mapping how one workflow really decides, ending in a written decision framework that is yours to keep regardless of what happens next.
- Hallucination
- When an AI model states something false with the same confidence as something true. Fine-tuning on your own documents and requiring evidence for every claim is how AutoServa keeps this from reaching a live decision unchecked.
- Segregation of duties
- A control that keeps the person or system preparing a transaction separate from the person or system approving it, so no single actor can complete a sensitive action alone. Workflow Guard enforces this for the operator too.
Still have a question?
Ask it on WhatsApp before you book anything. We answer honestly, including when the answer is that a workflow isn't a good fit yet.