Summary: Allin includes built-in guardrails to ensure it stays within its defined role, protects your brand, and does not engage in inappropriate or off-topic conversations. This article explains how guardrails work and how you can customise Allin's behaviour for specific scenarios using the Context Box.
Permissions Required: Location Admin or Brand Admin (to configure guardrail-related instructions)
What Are Allin Guardrails?
Guardrails are safety features built into Allin that ensure it:
Stays within its defined role and responsibilities as a hotel assistant.
Does not engage in conversations that could harm your brand or guests.
Will not perform or respond to irrelevant, unsafe, or malicious requests.
Gathers all necessary information before involving a staff member, reducing unnecessary back-and-forth.
How Guardrails Work in Practice
Intelligent information gathering — Allin's goal is to collect all the details needed to complete a task before escalating to staff. For example, if a guest asks to make a restaurant booking, Allin will ask follow-up questions (preferred time, number of guests, dietary requirements) so that when a staff member steps in, they have everything needed to confirm the reservation quickly.
Brand protection — Allin will not engage in off-topic conversations, respond to inappropriate requests, or say anything that could damage your brand reputation.
Automatic escalation — When Allin reaches the limit of what it can handle (e.g. before completing a booking or when it lacks information), it stops and escalates to staff.
Steps to Follow
To customise how Allin handles specific scenarios, add instructions in the Context Box:
Go to Settings in the left-hand side menu.
Click on the Location tab and select the desired property.
Open the AI Settings tab.
In the Context Box, add your specific handling instructions for the scenarios you want to customise. For example:
Sensitive issues: "If the message involves safety, security, or medical emergencies, respond with: 'I'm connecting you with a member of our team who can assist you right away.'"
Late check-in/out requests: "If a guest requests early check-in or late check-out, refer them to our late check-out policy."
Billing or payment issues: "If a guest asks about billing, respond with: 'For billing enquiries, please contact our front desk directly.'"
Complaints handling: "If a guest expresses dissatisfaction, acknowledge their concern, apologise, and let them know a manager will follow up."
Directing to external systems: You can instruct Allin to direct guests to a web link or booking system rather than escalating to staff.
Click Save.
Common Questions
Q: Can guests trick Allin into saying inappropriate things?
A: Allin has robust safeguards that prevent it from engaging in off-topic, inappropriate, or malicious conversations. These protections are continuously reviewed and updated.
Q: Will Allin make up information if it does not know the answer?
A: No. Allin only uses information it has been given access to. If it cannot answer, it will say so or escalate to a staff member.
Q: Can I turn off guardrails?
A: The core guardrails are built into Allin and cannot be disabled — they are there to protect your brand and guests. However, you can customise how Allin handles specific scenarios through the Context Box.
Q: Does Allin handle multiple guest questions at once?
A: Yes. Allin can respond to multiple guest questions in a row, drafting or automatically sending appropriate answers on behalf of staff.