Keen AI is the assistant integrated into the Keen Systems website chat. It connects a visitor's enquiry to an AI reply service while keeping a route to the team and WhatsApp. Its role is to support the conversation; a person still confirms project scope, commercial terms and decisions that need judgement.
What does Keen AI do on the website?
The website records a visitor's message in the chat conversation and queues a reply request. The integration sends the latest message with recent conversation context to the configured Keen AI service. When that service returns the expected reply decision, the website displays the response as Keen AI.
This is a documented internal website implementation. It is not a claim that every proposed AI use case has already been deployed for customers, or that the assistant has achieved a measured conversion improvement.
How does human handover work?
When a visitor requests a team member, the website pauses AI replies and marks the conversation for attention. Pending AI reply work is superseded. The chat also offers a route to continue on WhatsApp. Team availability and response time depend on actual staffing; a handover control does not promise that a person is always online.
The reply process checks the conversation again before posting an answer. If a staff member has taken over, the chat has closed or a newer message has arrived, an older AI response is not inserted as though it still belongs to the latest turn.
What happens if the AI service cannot reply?
The integration has a bounded retry path for selected temporary failures. If it cannot complete the reply, the chat tells the visitor that Keen AI could not answer and offers another message, a team-member request or WhatsApp continuation. The visitor should have an understandable next step when an automated reply is unavailable.
What information is used in a reply?
The website integration passes the current message, recent conversation history and the contact details collected for that chat to the configured AI service. The implementation sends up to the latest 30 conversation messages. Visitors should share only information needed for their enquiry and avoid passwords, payment credentials or unrelated sensitive records. See the privacy policy for the website's wider privacy information.
Conversation context helps the service follow the enquiry. It does not, by itself, establish how the upstream model was trained, which knowledge sources it uses or whether a particular reply is correct. Those are separate questions that need explicit configuration and evaluation.
Can a similar assistant be built for another business?
Yes, a comparable project can be scoped around a company's public information and support workflow. Start with approved questions and answers, define what the assistant may say, identify information requiring human review and make escalation visible. Connections to CRM, ERP, booking or order systems are additional work with their own access and validation rules.
- Public enquiries: explain services and help a visitor prepare a useful request.
- Staff escalation: preserve the conversation and pause automation when people take over.
- Business integration: plan authorised lookups or actions only after agreeing data ownership and permissions.
- Quality review: test representative questions, incorrect assumptions, unavailable services and handover behaviour.
Does Keen AI replace the sales or support team?
No. The implemented flow combines automated replies with a team-member request and WhatsApp continuation. Final quotations, commitments and exceptional support decisions remain matters for the team. An AI answer should be reviewed when accuracy is important, especially if it concerns a specific customer's records or a promised service.
How should an AI assistant pilot be evaluated?
Choose representative enquiries and record whether answers are supported, whether escalation works and whether the visitor can recover from a failed reply. Also test successive messages and a staff takeover while an AI response is still being generated. Agree evidence-based acceptance criteria before expanding scope; do not substitute a polished demo for operational testing.
Explore Keen AI, read the internal implementation case study or review our AI and ML development capabilities. To discuss a suitable assistant for your business, contact Keen Systems.