Chatbots and AI assistants
Custom Chatbot Development for Small Business
05 — Clinics & practices · shot in-house
What custom chatbot development means here
Custom chatbot development means commissioning a build rather than renting a tool. Arcline Digital designs, builds, tests, and launches an AI assistant that answers from your content, follows rules you set, and hands off to a human when it should. It is quoted as a project with a fixed scope. A build normally covers four stages: working out the questions your customers actually ask, preparing the source material the assistant answers from, building and testing the conversation logic, and connecting it to your website, inbox, or the systems you already run. You own what we deliver. There is no per-seat licence from us and no platform you are forced onto long term. We are a small remote studio working US and EU business hours, so you deal with the people writing the thing.
Thirty minutes, video or phone, US and EU hours.
What a chatbot build includes
The bot is the visible part. Most of the work sits behind it.
- A defined scope: the questions the assistant handles, and the ones it refuses
- A cleaned knowledge source built from your existing pages, documents, and FAQs
- Conversation design covering greetings, clarifying questions, dead ends, and escalation
- A handoff path to a human, whether that is an inbox, a form, a phone number, or a live chat tool you already pay for
- Testing against real customer questions before launch, not after
- Deployment on your site plus a short training session for whoever owns it internally
- Documentation of how to update answers when your business changes
The build process, stage by stage
Stage one
What should it do
We start with your actual enquiries. Support tickets, inbox threads, and the questions your team answers ten times a week. That list defines scope. A bot that tries to answer everything answers nothing well.
Stage two
The knowledge behind the answers
Most chatbot projects fail here. If your pricing lives in three places and two of them are wrong, the assistant repeats the error confidently. We audit the source material and fix the gaps, then structure it so the answers hold up.
Stage three
Build and internal testing
We build the assistant and put it through your own question list. Wrong answers get traced back to the source and fixed. We also decide what happens when it does not know something, which matters more than the impressive answers.
Stage four
Launch and tuning
It goes live where your customers already are. Then we review real conversations and adjust. After that, tuning either moves to a monthly care plan or to your team.
Who commissions this
- Businesses answering the same 20 questions by email every week
- Companies with a support inbox that has become a bottleneck for one person
- Service businesses that lose enquiries received outside business hours
- Teams evaluating an off-the-shelf chat tool and finding it cannot reach their booking system, catalogue, or internal data
If you need a generic chat widget and nothing more, an off-the-shelf product will be cheaper than us, and we will tell you so.
Off-the-shelf tool or custom build
An off-the-shelf chat product is faster to switch on and cheaper in month one. It also stops where its integrations stop, charges per seat or per conversation, and keeps your setup inside someone else's account. A custom build costs more up front and takes longer to scope. What you get back is control: the answers behave the way you decided, and the escalation path fits how your team actually works. Nothing gets renegotiated because a vendor changed its plans.
How the quote works
Chatbot work is quoted per project, in writing, before anything starts. The scope depends on how wide the assistant's remit is, how much source material needs cleaning up, whether it is deployed on your site alone or handed off to your team, and whether it integrates with systems you already run.
Two costs sit outside the project. AI provider usage is billed to you directly, so you see and control it. Ongoing changes are handled either by your team or by a monthly care plan, quoted per site.
Thirty minutes, and you leave with a scope and a number.
Scope boundaries
Not included in a standard build:
- Rewriting your entire website content library (we can quote that separately)
- Ongoing human support staffing
- Guaranteed answer accuracy on questions outside the agreed scope
- Integrations with systems that have no usable API or export
We also will not promise a chatbot replaces your support team. It removes repeat questions and covers the hours nobody is working. That is the honest version.
Get a scope and a number
Send two things: a link to your site, and the ten questions your team is tired of answering. We will come back with what an assistant could realistically handle and what it should not touch.
Thirty minutes, video or phone.
Studio hours cover US and EU time zones.
FAQ
How much does custom chatbot development cost?
Chatbot projects are quoted individually. Where a project lands depends on scope, how much source material needs cleaning up, and which systems it connects to. You get a fixed quote in writing before work starts. AI provider usage costs are billed to you directly rather than marked up.
Is this a product I subscribe to?
No. This is development work, delivered as a project. You commission a build against an agreed scope, and you own the result. There is no monthly fee to Arcline Digital for the assistant itself. Optional care plans exist if you want us handling updates and tuning, but they are separate and optional. If you prefer, your own team maintains it after handover.
How long does a chatbot build take?
Most builds go live in one week. A larger project has its core delivered in that week, and integrations can add time. The stage that stretches is usually knowledge preparation, not the build. If your answers are already documented and accurate, things move quickly. If pricing, policies, and service details are scattered across old PDFs that nobody has updated in years, that cleanup comes first. We tell you which situation you are in during scoping.
What happens when the chatbot does not know something?
It says so and hands the conversation to a person. That path is designed before launch, not bolted on afterwards. Depending on your setup, an unanswered question becomes an email to your inbox, a form submission, a callback request, or a transfer into a live chat tool you already use. Every escalation is logged so you can see what the assistant is missing.
Can it connect to the systems we already use?
Usually, yes, if the system has an API or a reliable export. Booking tools and CRMs are the common ones. Integration work widens the scope, because it needs testing against real data and real edge cases.
Schedule
Pick a slot, get a plan.
Thirty minutes, a clear plan and a fixed quote — video or phone, US and EU hours.
Book a call