AI Instructor Notice

Last updated: July 10, 2026.

Savi LoopIQ's AI Instructor is an educational assistant powered by a third-party large language model (currently Perplexity). It is designed to explain control-theory concepts and reason about the numbers it can see in your current tuning session. It is not a licensed engineer.

The AI Instructor can be wrong. Language models can produce fluent-sounding but incorrect explanations, misidentify the source of a phenomenon, or invent equations that look plausible. Never treat its output as engineering advice. Verify claims against a textbook, your control-theory references, and — for anything you plan to apply to real equipment — a qualified engineer.

1. What we send to the AI provider

When you send a message to the AI Instructor, we transmit to the AI provider:

Your prompt text.
Grounding context: current loop name, plant model parameters (K, tau, dead time), recommended gains, robustness readouts, and diagnostic flags for the loop you are viewing.
A system prompt establishing the assistant's role and safety guardrails.

We do not transmit raw uploaded CSV files, your other saved loops, or account metadata beyond the current session.

2. What the AI provider does with it

Perplexity processes the prompt to generate a response. Per our agreement with Perplexity and per our Privacy Policy, prompts submitted through Savi LoopIQ are not used to train third-party models. Perplexity may retain requests briefly for abuse detection and reliability.

3. Limits of the AI Instructor

It has no memory across chats unless you explicitly re-share context.
It does not access your historian, DCS, or any external systems.
It cannot verify that its numeric answers are dimensionally consistent or physically meaningful for your process.
It may occasionally cite references that do not exist or misattribute quotations.

4. Do not paste sensitive data

Because prompts are transmitted to a third party, do not paste into the Instructor any information you would not send in a normal business email — for example, ITAR/EAR-controlled technical data, personally-identifying information, or trade secrets you are not authorized to share.

5. Reporting a problematic answer

If the Instructor produces an answer that is dangerous, offensive, or clearly wrong in a way that could mislead a user, email safety@savagetechnicalsolutions.com with the prompt and the response text. We tune the system prompt and guardrails based on these reports.

6. Turning it off

You can ignore the Instructor at any time; nothing in the core tuning workflow requires it. Enterprise Team customers who want the Instructor disabled at the account level can request it via support@savagetechnicalsolutions.com.