The problem with most website chats
Many business websites have a chat widget. Often nobody answers it after office hours. Sometimes it collects a phone number and nothing else. Your sales team then calls people who only wanted a brochure, while the serious buyer who wrote at 11 pm has already contacted a competitor.
An AI assistant can fix this when it is set up with care. It replies at any hour, answers from your own information, asks the right questions and sends the promising conversations to a person quickly.
How it works
It answers from your information
The assistant is given a knowledge base made from your own material. This usually includes service pages, price lists or price ranges, FAQs, brochures, policies, project details and timings. When a visitor asks a question, the assistant finds the relevant part and writes a reply in plain language.
The key rule is that it should only answer from this material. If the answer is not there, it should say so and offer to connect the visitor with your team.
It asks qualifying questions
Once the visitor's question is answered, the assistant asks a few short questions. These depend on your business. Examples:
- Builder: preferred location, budget range, configuration, and when they plan to buy
- Clinic: the concern, preferred branch, and whether they want an appointment this week
- School: the grade sought, the academic year and the area they live in
- Manufacturer: product, quantity, delivery location and whether they need samples
- D2C brand: order number for support, or product interest and pin code for pre-sale queries
Keep it to three or four questions. The assistant should ask them naturally inside the conversation, one at a time.
It hands hot leads to a person
When the answers match what you define as a good lead, the assistant passes the conversation on. Common routes are a WhatsApp message to the right salesperson, a new lead in your CRM with the full chat attached, or an email to a shared inbox. The visitor is told clearly what happens next and when to expect contact.
Lower-priority enquiries can still be saved and followed up later, for example with a brochure on WhatsApp or an email.
What a good conversation looks like
Picture a visitor on a builder's website at night. They ask whether a 3 BHK is available in the project near the metro line. The assistant confirms the configurations on offer, shares the carpet area range and the RERA registration number from the project page, and mentions that a sample flat is open on weekends.
It then asks when they plan to buy and what budget range suits them. The visitor says within three months and gives a range that fits. The assistant asks for a name and a WhatsApp number, explains that a sales executive will message the next morning, and asks whether a site visit this Saturday would help.
By the time the executive opens WhatsApp, the lead arrives with the full chat, the budget, the timeline and the preferred visit slot. The first call starts with a useful conversation instead of the same basic questions.
Compare that with a form that only says Name, Phone, Message. The difference in quality of information is large, and the visitor got real answers too.
Setup steps
- Write down what a qualified lead means for you. Agree on this with your sales team before touching any software.
- Gather your source material. Collect current prices, policies and FAQs. Remove anything outdated. The assistant will repeat whatever you give it.
- List the questions it must never answer on its own. Final prices, discounts, medical advice, legal opinions and delivery promises are typical examples.
- Write the qualifying questions and the rules for scoring the answers.
- Connect the handover. Set up WhatsApp Business, your CRM or email so that leads arrive with the chat transcript and contact details.
- Add a consent line. Tell visitors they are chatting with an AI assistant and how their details will be used.
- Test with your own team. Ask awkward, vague and tricky questions. Fix the gaps in the knowledge base.
- Go live on a few pages first, such as your contact page and top service pages, then expand.
What to measure
- Conversations started and how many reach the qualifying questions
- Qualified leads handed over each week
- Response time from handover to a human reply
- Lead to meeting or sale, tracked in your CRM
- Unanswered questions, which show you what to add to the knowledge base
- Wrong or weak answers found during weekly review of transcripts
Read a sample of real conversations every week, at least in the first two months. It is the fastest way to improve the assistant.
Pitfalls to avoid
Wrong answers
AI models can produce confident replies that are wrong. Reduce this by limiting the assistant to your own material, telling it to admit when it does not know, and keeping that material current. When prices or policies change, update the knowledge base the same day.
No human handover
An assistant that traps people in a loop does more harm than having no chat at all. Always offer a clear way to reach a person. Make sure someone actually watches the handover channel during working hours.
Careless handling of personal data
Chats often contain names, phone numbers and sometimes health or financial details. Collect only what you need. Tell visitors why you are collecting it. Restrict who can see transcripts. Choose providers that explain where data is stored and how it is used. India's Digital Personal Data Protection Act, 2023 sets duties for businesses that handle personal data, so take advice on what applies to you.
Too many questions
A visitor who asked about parking does not want a form. Answer first. Qualify second.
A quick checklist before launch
- Lead definition agreed with sales
- Knowledge base current and approved
- Off-limits topics listed and tested
- Handover to WhatsApp or CRM tested end to end
- AI disclosure and consent text visible
- Named owner for weekly transcript review
How WebCrea can help
We build AI assistants that answer from your own information and pass serious leads to WhatsApp or your CRM. We also help with the less exciting parts, such as cleaning up source material and setting the rules for handover. If you want to explore whether it suits your business, we can talk it through.