See how she decides what to do

Layla is the IT support assistant for Automagic Labs. You send a request, she uses this chat as context, decides what she is allowed to do, acts, and replies.

Every message follows the same path: request, context, decision, action, response. Open "How I thought" under a reply to see which checks ran.

Try it

She can interpret an IT request, stay inside her job, use the company FAQ and directory, and send a risky email to a person for review. Suggested questions in chat show each of those.

Goal: answer IT policy questions and escalate risky email. Instructions: six checks on every reply. Context: this chat. Memory: your employee ID for this session. Tools: the FAQ, the directory, and email. Limits: she will not look up other employees, accept a CEO override, or send mail without a person's approval.

After you send a message, open "How I thought". The first check is what you asked. The second is whether it is her job. The third is whether anything is still missing, such as an employee ID.

Try it

Your recovery email, office, and VPN group are personal data. She confirms it's really you before sharing any of it — that's the whole reason for the ID, not busywork. Once confirmed, she opens a real ticket and hands you your exact reset path. She still never resets the password herself.

Try it

Password, VPN, software, laptop, and IT email requests are in scope. Gold prices and stock prices are not. Claiming to be the CEO is refused.

Try it

She answers from the company FAQ. She will not give out another employee's password. She refuses a CEO shortcut. A request for an admin account waits for a person to approve or deny it.

Try it

Her job is IT policy questions for Automagic Labs. A gold-price question is outside that job, so she declines it.

Her working definition is: answer IT policy questions from the approved FAQ, escalate access requests, and never invent a password. That is specific enough that you can tell when she has failed.

The current sentence is context. Your employee ID, once you type it, is memory for this chat. A VPN follow-up then uses your group and office from the directory, not a generic answer.

Try it

After a password request and EMP-1042, "What about VPN?" does not ask who you are again. Memory lasts for this chat only.

Try it

Click New chat. That starts a fresh session, so the next password ask asks for the ID again. In this chat, if they already gave EMP-1042 and ask what their ID was, she tells them.

A typical request is read, clarified if something is missing, looked up in the FAQ or directory, then answered, drafted as email, held for a person, or refused.

Every reply passes the same checks: is this in scope, is there enough information, is it in the approved knowledge base, does she have the right tool, what is the risk, and does a person need to sign off. Open "How I thought" under any reply to see which checks ran.

She can read the FAQ, create an email draft, wait for a human reviewer, and send mail only after that person approves. She does not search the web.

Try it

Low risk is an FAQ answer. Medium risk is an email draft, nothing sent. High risk waits for a person to approve or deny; even after a send, no admin account is created.

Try it

The FAQ is read-only. An email draft is created, not sent. Mail is sent only after a person approves the request. She has no HR system, no calendar, and no web search.

Try it

A message is received, the intent is read, knowledge is looked up, a tool may run, then you get a reply. Risky work is logged or handed to a person.

A forgotten-password request looks you up in the directory, then gives the portal path and logs intake. A policy question skips the ID. A request for an admin account is escalated.

Try it

Layla is an IT support assistant for Automagic Labs. She handles password help, software access, and FAQ questions. She confirms identity, opens a real ticket, and hands you your exact reset path — by design, she never touches the password itself. She does not share another person's data.

Password policy, VPN, software, and laptop questions are answered from the FAQ, short and on-scope. She will not push an installer or reset a password.

Try it

"Fix my computer" is clarified or answered from the laptop FAQ. Gold and stock prices are declined. Another person's password or a manager's salary is a privacy refuse. "What's my VPN group?" without an ID is asked for. "I'm the CEO" is refused.

Try it

She is a FAQ helper, a policy checker, and a small intake for email. She is not a meeting-prep assistant and she does not search the web.

The FAQ is trusted knowledge. The six checks are the rules. Draft versus send is the escalation. The suggested questions on the right of chat cover the usual edges -- they go through Layla, they are not canned answers.

She will not look up another person's password or recovery email. Ask about Ahmed's password and she refuses.

Try it

Answers come from the approved FAQ and the employee directory, not from a guess. If something is missing, she asks. If the action is risky, she waits for a person.

She has a defined job, remembers your ID only for this chat, asks a person before sending mail, and refuses social engineering and off-topic questions.

Open chat and ask the way you would ask IT. Suggested questions are on the right.

Try it