See the rules in real examples before using them
Calibration goes beyond a general instruction to use a friendly tone. Each rule is tested on typical and challenging messages.
Responses to praise and complaints
The hotel wants to respond briefly, address the specific point, sign off with the team name and not promise compensation publicly.
Rules and response samples
The panel shows how the configuration works on a historical compliment, question and complaint before it is passed to the tools.
- one review and comment configuration
- samples for different types of messages
- clear language and promise boundaries
A dedicated rules editor, not a loose description of tone
Calibration covers style, languages, sign-offs, required elements, prohibitions and situations requiring handover to a human.
Defines the language
Sets the tone, length, forms of address, sign-offs and response rules for different languages.
Defines the content
Specifies the required elements, permitted facts and promises the brand may use.
Sets boundaries
Records prohibitions, exceptions and situations in which a standard response should not be generated.
How is the response standard created?
History analysis
The module uses review and comment topics, along with audit results from previous responses.
Set the rules
The team defines the tone, length, languages, sign-offs, promises and exceptions.
Test with examples
AI creates samples for different types of historical messages.
Approve consciously
An authorised person approves the version passed to response suggestions and review.
Every rule has a version and test result
The configuration history shows which standard was used to prepare or assess a given response.
What does AI use?
- review topics and trends
- comment topics and momentum
- audit results for previous responses
- brand rules, languages and permitted promises
What does the team receive?
- versioned response configuration
- samples for different types of messages
- boundaries and escalation conditions
Rules take effect only after approval
AI can suggest improvements based on analysis and benchmarking, but it does not publish a new configuration independently.
An authorised user edits and approves the rules, checks the samples and decides when the new version takes effect.
The module does not respond to messages, publish content or automatically change the rules based on a single review or comment.
One definition of the brand voice
Employees and AI use the same rules for public communication.
Required elements and prohibited promises are clear before a response is prepared.
The team sees the effect of a change on examples before publishing the rules.
Calibration powers response suggestions and review
Analysis and audit results help build the rules, while additional sources and team interviews add to the brand knowledge base.
Related process elements
Frequently asked questions
Can different properties have different rules?
Yes. The configuration can apply to the entire brand, a specific hotel, a language or a type of message.
Can responses be tested before implementation?
Yes. The rules are checked against historical examples before publication.
Does AI change the brand tone by itself?
No. It can suggest a change, but an authorised user approves it.
Can calibration prohibit specific promises?
Yes. You can specify promises, compensation and wording that are not permitted in public responses.