AI Assistants
Website chat and lead qualification
A website chat assistant that answers from your approved facts, captures inquiries, and hands off to your team.
Visitors leave with unanswered questions.
Approved answers, inquiry capture, and a clear next step.
$300–$1,000 setup + $50–$150/mo Estimate
Hands off to a person and stays within approved answers.
Overview
How it works
Visitors often land on your website with one question. If they can't find the answer fast, they leave. Your website chat assistant answers that question in plain language, using only the facts you've approved, and gives the visitor a clear next step.
It says it's an AI if asked and never pretends to be a person. When it doesn't know something, it says so and offers to pass the question to your team. It doesn't guess, and it doesn't quote prices unless you've approved the exact wording.
When a visitor is ready to talk, the assistant collects their contact details and a short description of what they need. That inquiry lands in your dashboard and starts your follow-up, so the conversation keeps going after the chat window closes.
Urgent questions get special handling. If a visitor mentions something you've marked as urgent, the assistant points them to the right number and alerts your team.
What it looks like
A typical walkthrough
For example, here's how website chat works for a landscaping company on a busy spring weekend.
A visitor opens the chat
On Saturday evening, a homeowner asks whether the company handles weekly mowing in their town. The assistant answers from the approved service list and service area.
They ask about price
The visitor asks how much weekly mowing costs. The assistant explains that pricing is confirmed after the team learns about the property, or uses the approved price statement if the owner set one.
An unknown question gets an honest answer
The visitor asks about a plant the company has never discussed. The assistant says it isn't sure and offers to pass the question to the team.
The visitor leaves their details
The assistant asks for an email or phone number and a sentence about the job. It asks permission before anyone texts them.
The inquiry is saved
Once the system saves the inquiry, the assistant gives the visitor a reference and says the team will follow up. A confirmation goes out and the owner gets an alert.
An urgent message is routed
Another visitor types that a sprinkler line burst. The assistant shows the call-now number and alerts the staff member assigned to urgent issues.
Under the hood
How our system does it
Approved facts only
The assistant's knowledge comes from your business profile: services, hours, service area, and FAQs you've marked as approved. Starter FAQs stay switched off until you approve them, so nothing unreviewed reaches a visitor.
Rules it can't break
You list phrases and claims the assistant must never say. Every answer, whether written by the AI or by the offline system, is checked against those rules before it's shown.
A strict price policy
By default, no price is ever stated, even if one slips into an approved FAQ. If you choose, the assistant can use one approved price statement per service, word for word, and nothing else.
Urgent keywords route to a person
You choose the words that mean urgent. When a visitor uses one, the assistant shows the right number for the time of day and alerts the staff you've chosen, once per conversation. It says the team was alerted only when someone is actually set up to receive the alert.
Leads are captured with a tool call
When a visitor wants to start, the assistant uses a capture_lead tool. The system saves the inquiry only when there's an email or phone number and a description, and the assistant's confirmation is built from what was actually saved.
Protection against prompt injection
Visitor messages are treated as untrusted text. They're wrapped and cleaned before the AI sees them, they can't pick tools or change pricing rules, and the business they're talking to is set by the page, not by anything a visitor types.
Offline fallback answers
If the AI service is unavailable, refuses a request, or hits a daily limit you set, the assistant switches to fixed answers drawn from your approved FAQs. Visitors still get accurate answers and a way to reach you.
Powered by Claude
When AI is on, Claude, the AI model from Anthropic, phrases the answers in short, plain sentences. It's never used to decide prices, urgent routing, or bookings. Those follow fixed rules.
Works with
What it does
What it does
Answers visitor questions
Gives short, plain answers about your services, area, and hours from your approved facts.
Admits what it doesn't know
Says it isn't sure and offers a handoff instead of guessing.
Follows your price policy
Never states a price by default, or uses only your approved price wording.
Captures inquiries
Collects contact details and a short description, then saves the inquiry to your dashboard.
Asks before texting
Records text message consent only when the visitor clearly agrees, and mentions that STOP opts them out.
Routes urgent issues
Shows the right number for urgent problems and alerts your chosen staff.
Starts your follow-up
Each saved inquiry triggers a confirmation and an alert to your team.
What it can do for you
Why it helps
Answers outside business hours
Visitors get accurate answers in the evening and on weekends.
Clearer inquiries
You receive inquiries with contact details and the visitor's own description of the job.
Fewer repeat questions
Common questions get answered on the site instead of in your inbox.
Consistent answers
Every visitor hears the same approved information, in the same calm tone.
What you control
Your rules, your approvals
- Which FAQs are approved and live
- Phrases and claims the assistant must never say
- Your price policy: never quote, or approved statements only
- Urgent keywords, numbers, and who gets alerted
- How long chat transcripts are kept
- A daily limit on AI messages, and turning AI on or off
Who it's for
Is this a good fit?
- Businesses whose website visitors ask the same handful of questions
- Owners who want inquiries captured in the evening and on weekends
- Teams that need urgent requests to reach a person quickly
- Probably not the right fit if your visitors mostly need custom advice that can't be written down as approved answers
How we work
From first conversation to live
Discovery
We look at what visitors ask today, what you want them to do next, and which topics are off limits.
Setup on your business profile
We build your approved FAQs, services, hours, price policy, urgent keywords, and contact routing into one business profile.
Test in safe mode
We run test chats together, including tricky questions and attempts to break the rules, and adjust until the answers are right.
Go live
We add the chat widget to your site and watch the first conversations with you.
Ongoing review
We review transcripts, approve new FAQs with you, and tighten rules based on real questions.
Scope
What's included, and what isn't
Exactly what's included is confirmed in writing for each project.
Included
- Chat widget on your website
- Approved-FAQ knowledge built from your business profile
- Must-not-say rules and price policy checks on every answer
- Lead capture into your dashboard with a confirmation and team alert
- Urgent keyword routing and staff alerts
- Offline fallback answers when AI is unavailable
- Prompt-injection protection on visitor messages
- Testing in safe mode and ongoing transcript review
Not included
- AI usage and messaging charges from third-party providers, which are confirmed per project
- Writing your FAQs from scratch without your review and approval
- Live staff to monitor chats
- Answers on topics outside your approved knowledge
What you'll provide
- The questions visitors ask most, with answers you're happy to approve
- Your services, hours, and service area
- Anything the assistant must never say, written as clear rules
- Your price policy and any approved price statements
- Urgent keywords, call-now numbers, and who should be alerted
Safeguards
Controls built in
- The assistant says it's an AI and never claims to be a person
- It answers only from approved facts and hands off what it doesn't know
- Every answer is checked against your must-not-say rules and price policy
- It confirms an inquiry or alert only after the system completes it
- Visitor messages can't change tools, pricing rules, or which business is answering
- Text messages go only to visitors who clearly agreed, and STOP is honored
- You own your chat transcripts and lead data
FAQ
Questions owners ask
Will the chat assistant make things up?
No. It answers only from facts you've approved. When something isn't covered, it says it isn't sure and offers to pass the question to your team.
Can it give prices?
By default, it never states a price. If you prefer, it can use an approved price statement for a service, word for word. For Vyybe's own work, scope and price are confirmed in writing after a short conversation.
What happens when a visitor wants to talk to someone?
The assistant collects an email or phone number and a short description, saves the inquiry, and lets them know the team will follow up. You get an alert, and the visitor gets a confirmation.
Can a visitor trick it into saying something it shouldn't?
Visitor messages are treated as untrusted text. They can't choose tools, change price rules, or switch the business the assistant speaks for, and every reply is checked against your rules before it's shown.
What if the AI service goes down?
The assistant switches to fixed answers from your approved FAQs. Visitors still get accurate information and a way to reach you.
Does it text visitors?
Only if a visitor clearly agrees. Every text includes a way to opt out, and replying STOP ends texts right away.
Put website chat and lead qualification to work
Tell us what you need. We reply with a few questions, and scope and price are confirmed in writing before any work starts.