We confirm scope and price in writing. The main factors are:
- Scope of questions and languages.
- Knowledge preparation from your content.
- Platform and usage charges.
- Connections to email, CRM or WhatsApp.
- Testing and review.
Platform subscriptions and usage fees are paid to the provider and vary with traffic, so we show them as items to confirm. Our fee covers planning, setup, testing and handover.
A simple FAQ and lead-capture bot is a small project. A multilingual bot connected to a CRM with several handover paths is larger. We usually recommend starting with a narrow scope and widening it as real conversations show what visitors need.
Tell us about your site and the questions you receive, and we will reply with a plan and a quote.
Design matters more than many owners expect. A chatbot window that covers the contact button, opens without being asked or blocks content on a phone annoys visitors. We keep the widget small, place it where it helps, delay automatic pop-ups, and make the close button obvious. The first message should say what the bot can help with, so visitors do not guess.
Conversation design is a craft. Short answers read better than long ones, and a follow-up question such as which service the visitor is interested in keeps the exchange moving. We write replies in the tone of your brand, friendly and plain, and avoid jargon. We also plan what happens when the bot does not understand, with a polite request to rephrase and then an offer to connect with a person.
Speed and site health are checked too. Chat widgets load scripts, and a heavy script can slow a page. We choose lightweight tools where possible, load the widget after the main content, and test page speed before and after. If a tool noticeably slows your site, we say so and suggest an alternative.
Reporting shows whether the bot earns its place. Depending on the tool, you can see how many chats began, which questions were asked, how many ended in a lead or a handover and where the bot failed. We review these figures with you, and we avoid claiming results that the data does not show. Over time, the questions the bot could not answer become a useful guide to what your website should explain better.
Ownership and control are worth agreeing early. You should know who owns the knowledge base, where chat logs are stored, who can edit answers and how to switch the bot off quickly if something goes wrong. We document these points and give your team simple access to update answers when prices, hours or policies change, so the bot never repeats stale information.
Staff training is short but valuable. Team members learn how handovers arrive, how to respond quickly, how to mark a chat as resolved and how to report a wrong answer. A bot that passes a lead to an inbox nobody checks does more harm than no bot, so we agree who is responsible for each handover and what the target reply time is during working hours.
Some businesses worry that a chatbot will feel impersonal. In our experience the answer is design: a clear introduction that it is an automated assistant, honest limits, quick access to a person and a warm tone. Visitors generally accept a bot that is helpful and honest about what it is, and they dislike one that pretends to be human or traps them in a loop.
Finally, we plan improvements. After a month of conversations we group the questions, add new answers, remove unhelpful ones and adjust the greeting. After a few months we review whether the bot should take on new jobs, such as booking calls or qualifying leads with a few simple questions. Growth in small, tested steps keeps the quality high and the risk low.
Accessibility is part of good chatbot design. The widget should work with a keyboard, have readable text and sufficient contrast, and not rely on colour alone. Visitors using screen readers or older phones should still be able to reach your contact details, so the chatbot is always an addition to your contact page and never a replacement.
We also keep a simple record of the bot’s instructions, sources and version, so that anyone can see what it was told to do and when it changed. If a visitor reports a wrong answer, this record lets us trace the cause and correct it quickly.
