AI chatbot development

AI Chatbot Development

Chatbots we build

The difference between a chatbot people trust and one they close is governance: does it answer only from content you approved, say "I don't know" when it should, and hand off to a human with the conversation intact? That is how we built <a href="/projects/moby-1">Moby</a> — a governed assistant for hotels, vacation rentals, real estate, and auto dealerships that answers from each business's own property details, policies, and service information, 24/7, in 30+ languages, over chat and voice.

• Support chatbots that resolve the 60–80% of questions already answered somewhere in your docs
• Guest and visitor assistants for hotels, rentals, real estate, and dealerships — the Moby pattern
• Internal knowledge bots that answer staff questions from policies, wikis, and manuals
• Lead-capture bots that qualify visitors and book calls instead of showing a contact form

A chatbot engagement starts with your content, not a model: we inventory what the bot should know (help articles, PDFs, policies, product data, FAQs from real tickets), clean and structure it, and build retrieval that cites sources. Then we define the boundaries — topics it refuses, PII it redacts, when it escalates — and test against 200–1,000 real questions before anything goes live. Channels come last: a web widget first, then WhatsApp, SMS, Slack, email, or voice as needed.

When a chatbot needs to do things — look up an order, change a booking, update a CRM record — it becomes an agent, and we build that too: see AI agent development. Many engagements start as a grounded Q&A chatbot and add one or two actions once the answers are trusted. Pricing, token math, and ROI are in our AI chatbot development cost guide; for regulated content there is HIPAA-compliant AI development.

What AI chatbot development costs in 2026

Packaged assistant (support, FAQ, lead capture) configured on your content with one or two channels: from $4k, 3–5 weeks, roughly $150–$300 a month to run. Custom chatbot with your own retrieval, tone, boundaries, and integrations with a helpdesk or CRM for handoff: from $6k, 4–6 weeks, $200–$500 a month. Multilingual and voice (like Moby): add $2k–$5k, with voice at the top because of latency and interruption handling. Chatbots that take actions move into agent pricing at $12k–$25k. Every quote is fixed and written after a free discovery call.

• Content ingestion: help centers, PDFs, policies, product catalogs, CRM fields — cleaned, chunked, and cited.
• Guardrails: topic boundaries, PII redaction, injection defenses, refusal behavior, brand tone.
• Evaluation: a test set from real questions, run on every change, with accuracy and escalation rates reported.
• Human handoff: escalation into your helpdesk or inbox with the full conversation and the bot's confidence.
• Channels: web widget, WhatsApp, SMS, Slack, email, voice — in that order of cost.
• Analytics: what people ask, what the bot could not answer, and which content to fix next.

Chatbot vs AI agent: which do you need?

If the job is answering — "what is your cancellation policy", "does the room have parking", "which plan includes SSO" — you need a grounded chatbot, and it should ship in weeks for $4k–$8k. If the job is doing — issuing the refund, moving the booking, updating the record — you need an agent with scoped tools, approvals, and audit logs, which is a larger build. Most businesses are best served by starting with the chatbot, measuring deflection and the questions it cannot answer, and adding the first action once the answers are trusted. We build both, so the chatbot is designed to grow into the agent without a rewrite.