AI Chatbot

Site assistant that answers questions after hours, collects data and schedules contact. From the first sentence it says it's AI.

• individuallyScope
• AI and automationsArea
• individual quoteLaunch

Who this is for

Companies with a high volume of repetitive questions.

Problem

Clients write after hours and don't get answers. Companies with a high number of repetitive questions feel this the most — about price, hours, directions, availability. The same sentences have to be written for the hundredth time, and at eleven PM nobody writes them.

What we do

We place an AI chatbot on your site: answers questions, collects data and schedules contact. The assistant introduces itself as AI and doesn't pretend to be an employee — including if asked directly.

• Answers from your knowledge base, not general internet knowledge
• Collects contact data and saves it where you later look for it
• Schedules a contact or hands the conversation to a human when a question is beyond its scope

What's included

Three things that stay with you after implementation:

• Organized knowledge base and assistant response boundaries
• Conversation, data collection and contact handover scenarios
• Website chatbot, tests and user manual

Timeline

We determine the timeline individually and give it before launch, not during. The reason is one: work scales with knowledge base size and number of languages and integrations, and the base materials are on your side. The order is fixed though: knowledge base, configuration, tests, implementation.

Cost

The largest part of the work is usually not the bot itself but putting in order what it is supposed to know. On top of that come the integrations and the number of languages it has to answer in.

What you get

24/7 support and zero lost inquiries: a question asked at night gets answered at night, not the next afternoon. Repetitive questions stop consuming the team, and you get a record of what people really ask — including questions you don't have answers for in your offer yet.

How to calculate the value of one conversation caught after hours

The chatbot answers when nobody is at the desk, so its value can be calculated from your own data, not from promises:

• Value of a week of after-hours conversations = number of website visits outside working hours per week × share of such visits that would end with a left contact if someone answered right away (calculated on visits during working hours, where a human answers) × average value of one contact from your CRM

The middle factor comes from your own data, calculated on visits during working hours — it's the only number in the formula that can't be guessed, and without it the formula stays just a scheme.

How an AI chatbot differs from a contact form and a widget with preset buttons

Three ways of handling after-hours questions look similar on a website, but differ in what happens to a question nobody answers right away:

• Contact form — collects the question and waits until someone reads it the next day; the customer gets no answer at that moment
• Widget with preset buttons — follows a rigid script and only answers questions someone anticipated in advance
• Live chat with a human — answers right away, but only when someone happens to be online on the other end
• AI chatbot — understands a question asked in free text and answers from your knowledge base immediately, at any hour

The difference isn't cosmetic: the first two solutions only register the question, the third depends on a human being present, and only the fourth answers right away regardless of the time.

Where the chatbot uses a language model, and where it runs on plain rules

This distinction matters in practice, not just in terminology.

• The language model understands a question asked in free text and formulates an answer from your knowledge base — this is the only part that can honestly be called artificial intelligence
• The decision on when to ask for contact details and when to hand the conversation to a human is thresholds and rules set in advance together with you — plain conditional logic
• Saving the conversation into the system you point to is automation and a template, not reasoning

Calling the second and third point 'artificial intelligence' sounds more impressive, but isn't true — and it's exactly those two that decide when the customer reaches a live person.

When an AI chatbot doesn't pay off

There are situations where this layer won't help, and it's more honest to say so directly than to implement it and be disappointed a month later.

• The website gets few visits, and customer questions are different every time — there's no repeatability for a knowledge base to build on
• The offer and prices change daily, and nobody in the company has time to update the knowledge base — the bot will start answering with outdated information
• Most questions require an individual quote or a legal or medical assessment that can't be captured in a fixed set of answers
• Nobody in the company is ready to approve the boundaries of the bot's answers before launch — without that, the knowledge base has no owner

What you need to prepare before the start

Work on the knowledge base starts with materials nobody can invent for you:

• A written-down list of services, pricing and answers to the most common customer questions
• A place for collected contact details to land — a CRM or an inbox someone actually checks
• A decision-maker who will approve the bot's answer boundaries before it goes live on the website
• Agreement that the assistant identifies itself as AI from the very first line — this isn't up for negotiation

How to check after a month whether the chatbot paid off

Metrics are worth setting before the start, so that after a month you're judging numbers from your own systems, not an impression:

• Share of conversations ending with left contact details — from the bot's conversation log
• Number of questions the bot couldn't find an answer to in the knowledge base — shows what's missing from it
• Number of conversations handed to a human, and how long it took someone to pick them up
• Number of after-hours inquiries before and after implementation, for the same period of the year

What the AI chatbot doesn't do

It answers from the knowledge base and collects data, but doesn't replace decisions nobody should hand to an automaton. It doesn't negotiate price beyond what's already in the price list, doesn't assess legal or medical matters, and doesn't sign any commitments on your behalf. Any conversation that goes beyond its scope can be handed to a human — a bot without that exit isn't automation, it's a wall the customer bounces off and walks away to a competitor.

The most common mistake when implementing a chatbot, and whose data stays with you

The most expensive mistake isn't a poorly configured opening message, it's an outdated knowledge base: prices change, nobody tells the bot, and it keeps answering with old figures until a customer catches it live. The second mistake is turning off the handoff to a human 'to keep things simple' — a hard question then goes unanswered instead of reaching someone who could handle it.

Chat history and collected contact details go into the system you point to — a CRM or an inbox — and stay there. It's your data, and you decide who has access to it.

What you get

• Structured knowledge base and clear answer boundaries
• Conversation, data-capture and human-handoff scenarios
• Embedded website chatbot, tests and operating instructions

When the problem lies elsewhere

If your problem sounds different, neighbouring areas of the same system stand right next to it — AURA connects them to each other rather than selling them separately:

• Request automation — Form goes to CRM, notification to the right person, and task to the calendar. No manual retyping.
• AI lead qualification — AI evaluates and tags inquiries by potential before anyone starts calling back. The team starts with the best inquiries, not the newest.
• AI follow-up — Message sequences for those who didn't respond. Second and third reminder go out on their own, at set time and with stop condition.
• CRM and automations — Requests, reminders and client data work on their own, without re-entry from paper to calendar.