A business chatbot should make it easier to understand your offer and start a useful conversation. EidenTech develops website assistants around approved company information, clear lead requirements, and the tools your team already uses. Our own lead assistant is a working example of this approach.
AI chatbot development services, from answers to follow-up
Choose the job the assistant must complete before choosing a model. A lead conversation, document search, and a customer account action have different knowledge, integration, and testing requirements.
- Approved business information
- Company-specific conversation
- Validated lead details
- Spreadsheet or CRM handoff
Website lead qualification
Explain approved services, clarify what the visitor wants to build, collect a name and contact method, and pass useful project context to the team. Our own website assistant demonstrates this flow.
Customer support and FAQs
Help customers find approved product information, policies, and troubleshooting steps. Define when the assistant should hand the question to a person, and keep the content current.
Document and knowledge assistants
Scope retrieval over an approved document collection with source references, freshness rules, and user access boundaries. This is an additional architecture decision, not a feature implied by every chat widget.
Connected business conversations
Plan supported CRM updates, booking requests, or order information against real APIs. Read-only answers and actions that change business records need separate permissions and acceptance checks.
How we approach your project
Define the conversation and the company knowledge
We begin with your services, audience, common planning questions, and the action that a helpful conversation should support. Approved company information establishes what the assistant may explain; it should not invent project results, prices, or capabilities. The initial scope can use a carefully maintained knowledge prompt. Larger or frequently changing document collections may call for retrieval, source references, and a content update workflow. Those are separate requirements rather than features that every chatbot automatically includes. We agree how the assistant redirects unrelated questions and handles requests that need a person’s judgement.
Turn interest into a usable lead
A conversation becomes useful to the team when it contains reliable contact information and enough context to follow up. We define the fields, validation, consent language, and follow-up questions before building the flow. EidenTech’s assistant asks for a name, email or phone, project interest, and timing. It only treats a record as ready to save when there is a name and a valid contact method; it can continue collecting the remaining project details. The application validates the data on the server and updates the same lead record during the session instead of relying on the model alone to decide whether a contact is complete.
Connect the assistant to your workflow
The website widget sends requests to a server API so model and integration credentials stay outside the browser. Our current implementation connects Groq and Gemini generation with provider-aware request timeouts and cooldowns, plus Google Sheets lead storage through a service account. If model generation is unavailable, a scoped fallback can still explain the company and ask for the next required detail. Your project can instead connect to a CRM or another supported business tool. We define field mappings, supported actions, failure feedback, and ownership of the provider accounts before connecting live customer data.
Evaluate and operate the chatbot
We test normal enquiries, missing contact details, unrelated requests, malformed input, provider failures, and duplicate lead updates. Prompt instructions alone do not establish a complete access or data validation boundary. Server-side controls, provider usage limits, and storage behaviour must be reviewed together. The appropriate rate-limit and session design depends on expected traffic and deployment; an in-memory prototype limiter may need a shared store for a larger multi-instance deployment. Launch planning covers knowledge updates, failure monitoring, follow-up ownership, and operating costs. A chatbot supports your sales process; conversion improvements need measurement with real visitors after release.
What we can deliver
We agree the requirements and responsibilities before development begins. Your project scope can include:
- A company knowledge and conversation brief with approved answers, boundaries, and lead requirements.
- An accessible website chat interface and server-side generation API with provider configuration kept private.
- Validated name and contact capture, progressive project questions, and a supported lead-storage or CRM integration.
- Representative conversation checks, failure handling, provider usage planning, and handover for knowledge updates.
Hosted widget or custom chatbot: what fits?
Compare what the release must accomplish, the integrations it needs, and who will operate it. These are planning options, rather than fixed packages or price quotes.
What affects development and operating cost?
A useful quote separates the initial build, recurring services, assumptions, and exclusions. Bring these four decisions to the first conversation.
- Knowledge and content
- A stable approved FAQ is narrower than indexing changing documents with access rules, citations, and an update workflow.
- Conversation and action scope
- Capturing an enquiry differs from authentication, multilingual flows, bookings, and actions that modify a customer record.
- Integration behaviour
- Field mapping, create/update rules, duplicate handling, error recovery, and sandbox availability determine integration effort.
- Operation and handover
- Estimate hosting, model usage, monitoring, evaluation, and maintenance separately. Free provider allowances can change and do not guarantee zero ongoing cost.
Acceptance checks for your first release
Use observable checks to assess the implementation and handover. The final test plan follows the agreed scope and the systems your business uses.
- Answers stay consistent with approved company information and route unsupported questions to the agreed next step.
- A record is saved only when the required name and valid contact method are present; incomplete details prompt a useful follow-up.
- The visitor sees a save confirmation only after the spreadsheet or CRM accepts the record.
- A continued conversation updates the intended enquiry rather than creating a new lead on every message.
- Provider timeouts and unavailable integrations have understandable recovery behaviour.
- The widget works with a keyboard, on mobile, and alongside a contact method outside chat.
Technology choices for your requirements
The stack follows your integrations, operating model, and maintenance needs. These are options to review during discovery, rather than a fixed stack promised for every project.
- Next.js, React and TypeScript website components
- Server-side Groq and Gemini generation APIs
- Google Sheets API, OAuth service accounts and validated lead records
Explore the assistant on this website
Our company assistant demonstrates approved business answers, progressive project questions, and a Google Sheets lead handoff. The application requires a name and valid email or phone before it saves an enquiry.
Read the implementation story for the flow, validation, failure handling, and current limitations. Opening the assistant stages a project question; it does not send a message or save a lead.
Plan your chatbot’s job in three decisions
Choose the conversation goal, approved information, and handoff. This planner creates a starting brief you can copy and discuss with us. It works in your browser without collecting contact details.
Planning questions, answered
Can you build a chatbot for my business website?
Yes. We can plan and build a website assistant around your company information and customer journey. Share your services, common questions, required lead fields, and the system your team uses to follow up.
Can the chatbot capture leads in Google Sheets or our CRM?
Yes. Our own assistant supports Google Sheets through a server-side service account and validates a name plus email or phone before saving. CRM integration depends on the provider’s supported API, permissions, and agreed field mapping.
Can the chatbot use our documents and company information?
Yes. A small, stable knowledge set can be included in the assistant’s approved context. A larger document collection may need a retrieval workflow, source references, permissions, and a process for keeping information current.
What happens when an AI provider is rate limited?
The implementation can use request timeouts, provider-aware cooldowns, supported failover, and a bounded fallback response. Provider limits and account terms still apply; changing keys does not create unlimited free usage.
How much does custom chatbot development cost?
Cost depends on the knowledge scope, conversation complexity, integration work, evaluation, and operating requirements. We first agree what the chatbot must do and which tools it needs to connect, then prepare a scope-based proposal.
Practical guides for this decision
How to Choose a Chatbot Development Company
Use a concrete business brief, a realistic demonstration and clear delivery terms to compare chatbot development partners before you commit.
Read the planning guide ↗CHATBOT PLANNINGHow much does an AI chatbot cost for a business?
A practical way to assess chatbot development and operating costs without treating a generic price range as a project quote.
Read the planning guide ↗Chatbot integrationChatbot CRM Integration: From Conversation to a Useful Lead
Plan the fields, validation, permissions and failure recovery that turn a chatbot conversation into a record your sales team can act on.
Read the planning guide ↗Chatbot launch planningAI Chatbot Launch Testing Checklist for Your Business
Review answer quality, lead capture, business permissions, failure handling and the customer experience before a website chatbot goes live.
Read the planning guide ↗Tell us what needs to work better.
Share your idea, existing workflow, and the outcome you need. We will help shape a practical starting point.
Start a project conversation ↗Read our website, chatbot, and product planning guides →