How is this different from a generic AI assistant on our website?

A generic assistant answers from general knowledge and guesses at your specifics. This one is built over your documentation, policies and FAQs, so it answers with your terms, your pricing rules and your process — and where it has nothing to answer from, it escalates instead of inventing something.

What does it need from us to get started?

The material you would give a new support hire: product documentation, FAQs, policies, and ideally an export of past support conversations so we can see what people really ask. Where that material is thin, the first weeks of transcripts usually show exactly what to write next.

What happens when the bot does not know the answer?

It follows the escalation rules you set — hand to a human, open a ticket, or take contact details for a callback — and it passes the conversation so far along so the customer does not repeat themselves. We would rather it escalate too often in week one than answer confidently and wrongly.

Can it capture leads and write them into our CRM?

Yes. Capturing the name, contact details and qualifying answers inside the conversation, then creating or updating the record in HubSpot or Salesforce, is one of the most common reasons businesses ask for this. Which fields and which system are agreed during scoping.

Can it run on WhatsApp or Slack as well as the website?

Additional channels are configured per deployment rather than switched on by default. We agree which ones are in scope before the build, because each one has its own approval process and its own constraints.

Will it sound like our business?

Tone is part of the build. We configure it against your existing written support replies and marketing copy, and you review real sample conversations before it goes live.

How do we know what it is telling customers?

Every conversation is available as a transcript, and common-question reporting shows what visitors ask most. In the first month that reporting is usually as valuable as the automation itself, because it shows which pages of your site are not answering the question.

Does it replace our support team?

No. It takes the repetitive first-line volume so your team spends its time on the conversations that need judgement, and the ones that reach them arrive with the context already gathered.

AI Chatbots

A website conversation that ends in an answer, not a form

Flasin chatbots are built on your own documentation and business rules. They answer what visitors actually ask, capture the details your team needs, take the next action where they can, and hand over to a person when they should.

Get Started Free AI Audit

Answers from your approved content
Captures and routes real leads
Escalates with the full conversation

Example conversation

Illustration, not a transcript

Visitor:
Does your onboarding plan include migrating our existing data?
Bot:
Yes — data migration is included in the onboarding package, covering up to two source systems.
Source: Onboarding Guide — Section 3
Visitor:
We have three. Can someone confirm what that would cost?
Bot:
That one needs a person. If you leave a name and email I'll pass this conversation to the team and they'll come back with a figure.

Lead captured in the CRM · conversation handed to a human with full history

Illustration of a conversation the chatbot is configured to handle.

What it does

The difference between a chat widget and a chatbot that knows things

A generic assistant bolted onto a website gives generic answers, and visitors work that out in two messages. What makes a chatbot worth having is what it has been given to work from: your product documentation, your policies, your pricing rules, your FAQ, the answers your support team already types every day.

Built that way, a conversation can finish. The visitor asks whether you support their setup and gets a real answer with a link. They ask what a plan includes and get your terms rather than an invented summary. They ask for a callback and the contact record exists in your CRM before they close the tab.

And when the question is one only a person should answer, the bot says so and hands the conversation over — with everything already said attached, so nobody has to start again.

What a conversation can end in

A direct answer, drawn from your approved documentation
A qualified lead written into your CRM
A booked call or a scheduled callback
A support ticket opened with the context attached
A handover to a human, mid-conversation, with full history
An honest 'we should get a person to answer that'

Which outcomes are in scope is agreed during discovery. A bot that cannot do any of them is a search box with a personality.

Capabilities

What the chatbot handles

Each of these is configured against your content and your rules — not a switch on a platform.

Website conversations

Answers visitors in natural conversation on your site, at any hour, including the messages that arrive at 11pm on a Sunday.

Answers from your knowledge base

Responses are grounded in your documentation, policies and FAQs, so the bot says what your business says rather than what a general model assumes.

First-line customer support

Resolves the repeat questions your team answers dozens of times a week, and collects what a specialist would ask for before escalating the rest.

Lead qualification

Asks the questions that decide whether an enquiry is a fit, and records the answers against the visitor's record.

Lead capture

Collects name, contact details and intent inside the conversation instead of pushing the visitor to a separate form.

Booking and action routing

Books a call, opens a ticket, or triggers the workflow the conversation calls for, rather than promising that somebody will be in touch.

Human escalation

Hands the conversation to your team on request, on low confidence, or on any topic you mark as human-only — with the full transcript attached.

Deployment where your customers are

Website first, with messaging channels such as WhatsApp, Telegram or Slack configured per deployment where they are part of your scope.

Conversation review

Transcripts and common-question reporting show what visitors actually ask, which is usually the most useful output in the first month.

Channels, integrations and escalation rules are scoped before the build starts. We would rather tell you a channel is out of scope than list it here and quietly not build it.

How it works

Visitor → chatbot → your knowledge and rules → answer, capture or route

The bot is a front door onto material you approved and actions you defined. Nothing in the middle is left to a general-purpose model's judgement.

Website visitor

Somebody with a question, on your site, right now.

Product or service question
Pricing or fit enquiry
Support request

Flasin chatbot

Works out what is being asked and what it needs to answer it.

Understand the question
Ask for what is missing
Stay inside scope

Knowledge & business logic

Your documentation and your rules decide the response.

Approved content
Qualification rules
Human-only topics

Answer, capture or route

The conversation ends in something, not in a promise.

Answer with a source
Lead into the CRM
Ticket, booking or handover

Where the bot is not confident, or the topic is one you have marked as human-only, it escalates rather than answering. That rule is set by you, not inferred.

Business use cases

The conversations businesses hand over first

All of them share a shape: high volume, repeatable answers, and a next step that should happen while the visitor is still on the page.

E-commerce

Order, delivery and returns questions

Answers where-is-my-order, delivery and returns-policy questions from your published terms, and escalates the exceptions to your support team.

SaaS

Product questions and onboarding

Answers setup and feature questions from your documentation, and points a trialling user at the exact page rather than the docs homepage.

Professional services

Enquiry screening

Establishes what the prospective client needs, whether it is something you take on, and books the consultation when it is.

Clinics & practices

Pre-appointment questions

Answers questions about services, hours and what to bring, collects intake details, and routes anything clinical to staff.

Real estate

Listing enquiries

Answers questions about a property, qualifies buyer or tenant intent, and books viewings into the agent's calendar.

Internal teams

Employee helpdesk

Answers HR, IT and policy questions from your approved internal documents, so the same question is not asked in three channels.

Marketing

Campaign landing pages

Qualifies paid traffic in conversation rather than through a five-field form, and hands the qualified ones straight to sales.

Education & membership

Programme and enrolment questions

Answers questions about programmes, requirements and dates from your published material, and routes applications to the right team.

Integrations

A conversation is only finished when your systems know about it

Answering well is half of it. The other half is that the lead, the ticket and the booking exist afterwards without anyone re-typing them.

CRM & sales

Create the contact, log the conversation, score and route the lead.

HubSpot
Salesforce

Support & helpdesk

Open the ticket, attach the transcript, escalate to a person with the history intact.

Zendesk
Intercom

Messaging channels

Take the same conversation to the channels your customers already use.

WhatsApp
Telegram
Slack

Anything with an API

Order lookups, membership checks, internal portals — where your system exposes an API the bot can read from it and write to it. Where it does not, we tell you during scoping instead of after.

Named systems are integrated per deployment using their public APIs. Flasin is not affiliated with, or endorsed by, any of them.

Implementation

How a chatbot gets built

Five stages, in this order. The first one is where most of the value is decided.

01
Understand
We read your existing support conversations and FAQs to find what visitors genuinely ask, and agree which of it the bot should handle.
02
Build
We build the knowledge base from your approved documentation, write the conversation and qualification logic, and connect the CRM, helpdesk and booking systems.
03
Test
We run the bot against real questions — including the vague, the hostile and the out-of-scope — and tune until the answers and the handovers are right.
04
Launch
It goes live on your site, your team gets the handover path and the transcript view, and we watch the first weeks of real conversations closely.
05
Improve
Real transcripts show the gaps in your documentation as much as in the bot. We review both with you and update on an agreed cadence.

Timelines depend on how much documentation exists, how clean it is, and which integrations are in scope. You get a timeline after discovery, not from a web page.

Security & oversight

A bot speaking for your business needs limits

Most of the risk in a customer-facing chatbot is answering confidently outside what it knows. These are the boundaries we build in.

Grounded in approved content

Answers come from the material you approved. The bot is built to say it does not know rather than to fill a gap convincingly.

Escalation always available

A visitor can reach a person at any point, and topics you mark as human-only never get an automated answer.

Full conversation history

Transcripts mean you can review what was actually said to a customer, not just what the bot was configured to say.

Least-privilege integrations

CRM and helpdesk connections request the narrowest access the workflow needs, rather than blanket account-wide permissions.

Consent-aware deployment

The widget respects the site's cookie and consent settings, and what it stores about a visitor is agreed with you before launch.

Retention decided with you

How long transcripts and captured details are kept is a decision we make with you during implementation, not a default you inherit.

These are the practices we build to, not certifications we hold. Flasin has no SOC 2 report, ISO certification or HIPAA attestation. If your business requires a specific compliance framework, raise it in discovery and we will tell you plainly what is and is not achievable within scope.

Questions

What people ask before they start

Get started

Give your website visitors an answer instead of a contact form

Send us the questions your team answers most often and we will tell you which of them a chatbot should handle, which it should not, and what it would take to build.