AI Chatbots & Assistants
Answers in seconds, around the clock, without the robot embarrassment.
04.2 / Service
AI Chatbots & Assistants
In short
What is an AI chatbot for business?
An AI chatbot is an assistant that answers customer questions on your website or messaging channels from a knowledge base built on your real materials. We decide what it should and should not handle, deploy it on the website, Messenger or WhatsApp, and set the rules for handing over to a person. You get an approved knowledge base, a lead capture flow and a monthly conversation report.
The problem
The problem this solves
Inquiries arrive at midnight and on weekends; answers come Monday afternoon, and by then the customer bought elsewhere. Meanwhile most chatbots on the market frustrate more than they help, because they were deployed, not designed.
Method
How we work
Scope first: a bot that does three things reliably beats one that does thirty badly. It answers from your approved knowledge base, says honestly when it does not know, hands over to a human cleanly, and every month real conversations tune it sharper.
When it fits
When this is the right call
Half your inquiries arrive on Instagram after ten at night, and the first reply comes when someone opens a laptop next morning.
Your team answers the same twelve questions about delivery, sizes and opening hours every day, and real conversations wait behind them.
You tried a bot before. It answered everything with a menu, customers typed 'operator' and it was switched off within a month.
Leads from chat end up in screenshots and personal phones, and nobody can say how many became customers.
04.2 / In one line
Sales and support bots with human hand-off
Scope
What's included
Use-case design: what the bot should and should not handle
Knowledge base built from your real materials
Website, Messenger and WhatsApp deployment
Human hand-off rules and lead capture
Brand voice, bilingual responses, honest AI disclosure
Conversation analytics and monthly tuning
Scope shown is indicative: the final scope is agreed with you before work starts.
Deliverables
What you receive
Use-case map and out-of-scope list
The questions the bot answers, the ones it hands over and the ones it must refuse, agreed before configuration starts.
Approved knowledge base
Your price lists, policies and FAQs rewritten into short, answerable entries, each with an owner and a review date.
Conversation design and tone sheet
Opening lines, fallback phrases, AI disclosure wording and the exact sentence used when the bot does not know, in both languages.
Hand-off and lead capture flow
Rules for when a human takes over, where the transcript lands and which fields reach your CRM or inbox with the lead.
Pilot test log
Real conversations reviewed line by line, every weak or wrong answer logged, corrected and re-tested before launch.
Monthly conversation report
What people asked, what the bot could not answer, where it handed over and which knowledge gaps to close next.
Process
01 / 04
Scope
The questions worth automating, chosen from real inquiry data.
02 / 04
Build
Knowledge base, tone, guardrails and hand-off logic.
03 / 04
Pilot
Live on limited traffic, watched closely, corrected fast.
04 / 04
Scale
Full deployment with monthly conversation reviews.
Timeline
A first engagement, step by step
Days 1-3
We read your recent inquiries and group them. You see the list of what the bot will and will not handle, and approve it.
Weeks 1-2
Knowledge base written from your materials, tone agreed, hand-off rules configured. Your team tries to break it; we log what breaks.
Weeks 3-4
Pilot on one channel with limited traffic. A named person on your side watches hand-offs; we review every transcript and correct answers the same day.
Month 2
Full deployment on the agreed channels. The first monthly review shows what was asked, answered and refused, and what to add next.
Measurement
What we measure
Resolved without hand-off
The share of conversations the bot closed correctly on its own, judged by transcript review, not by the customer giving up.
Wrong-answer rate
Answers the review found false or misleading, per hundred conversations. This number decides whether the bot stays live.
Time to human
How long a customer waits after the bot hands over. A fast bot in front of a slow queue only moves the delay.
Leads captured with complete details
Chats that ended with a name, contact and need recorded in your CRM, so the bot is judged on customers, not message counts.
Proof
Evidence, not adjectives
FAQ
Frequently asked questions
Will the bot say wrong things to customers?
AI chatbots can make mistakes. We configure the bot to use approved content, decline unsupported requests and hand conversations to a person when needed. Testing and monthly conversation reviews help us identify and improve weak answers.
Do customers know they're talking to AI?
Yes, we disclose it clearly. Pretending a bot is a person damages trust and is exactly the kind of shortcut that backfires publicly.
Which channels can it live on?
Website chat, Facebook Messenger, WhatsApp and Instagram DMs, usually starting with the one where your inquiries actually arrive.
How is this priced?
Setup is a project: scope, knowledge base, channels and the pilot. After launch a monthly retainer covers hosting, model usage, transcript reviews and tuning. The price moves with the number of channels and the size and churn of the knowledge base.
What do I need to provide, and when is it live?
Access to the channels, your real inquiry history, current prices and policies, and one person who answers our questions within a day. A pilot usually runs by the third week; launch follows a clean pilot, not a calendar date.
Can't I just connect ChatGPT to the site myself?
A basic integration is possible, but a useful business assistant also needs verified business information, clear limits, testing and a hand-off process. These are the parts we design and configure.
Where to next
Keep going
Related services
Articles on this