Alactic ACID
Test the assistant, understand AI consumption and tune responses for useful business conversations.
Evaluate usefulness, factual consistency, handover and consumption together. A shorter answer is valuable when it still answers the customer's actual question.
Understand Alactic ACID's role
Know how the conversational intelligence system supports a business workflow.
Open the Test AI workspace
Try customer questions before changing a live conversation.
Test a complete customer conversation
Evaluate context across several turns rather than one isolated answer.
Check input and output AI usage
Understand why both the question and the generated reply affect consumption.
Read the workspace AI allowance
Use the actual displayed limit rather than a universal message estimate.
Distinguish AI usage from WhatsApp charges
Keep generation consumption separate from message delivery costs.
Use shorter answers where appropriate
Reduce unnecessary text while preserving useful information.
Reduce unnecessary knowledge in a test
Focus the assistant on relevant verified information.
Investigate an AI limit response
Explain why a request cannot run and what the user can do next.
Plan AI testing within an allowance
Prioritise representative tests instead of repeated open-ended chatting.
Compare response settings with evidence
Choose a configuration using repeatable examples.
Test factual accuracy
Check answers against the business's approved information.
Test uncertainty handling
Check how the assistant responds when the answer is not available.
Test resistance to instruction changes
Check that customer messages do not easily override business boundaries.
Test human escalation behaviour
Ensure a customer can move from automation to a person.
Review AI replies for tone
Check that language fits the situation as well as the brand.
Test multilingual AI responses
Evaluate supported customer languages with real business terms.
Keep sensitive information out of AI tests
Evaluate behaviour using fictional or minimised data.
Review an unexpected AI answer
Diagnose whether the cause is knowledge, settings or the question.
Understand shared capacity and workspace limits
Distinguish a workspace allowance from service-wide availability.
Plan a human fallback for AI unavailability
Keep customer support usable when generation cannot run.
Read the ACID information page accurately
Explain the product without overstating its technical origin.
Evaluate AI before a wider rollout
Use a structured release check for business conversations.
Monitor AI after knowledge changes
Check whether a content update changes answer quality or consumption.
Create an AI improvement backlog
Turn observed failures into prioritised product and knowledge work.
Relevant public sources
- Alactic Communications Terms & ConditionsAlactic public terms
- WhatsApp Business Messaging PolicyProvider policy