Aumiqx Technologies
AUM
Open menu

AI chatbot development services

From $2,500

AI chatbot development: a bot that knows your business, and knows when to stop.

We build chatbots and voice agents that answer from your own documents, show the page each answer came from, say “I don't know” when the answer isn't there, and hand the conversation to a person when it should. On your website, WhatsApp, Slack, Teams or the phone.

See what chatbots cost

By Axit · Updated Sep 24, 2026 · Prices from 7 agencies, 7 cost guides and 12 vendor pages, all linked

Try it: a chatbot trained on your website in seconds

Paste your address. It reads a few of your public pages and answers questions from them only, with the page it used. Ask something your site doesn't say and watch it refuse. Ask for a refund and watch it hand off. That is the whole job of a good chatbot, in miniature.

chatbot · paste a website to start live

It reads your homepage and up to two more public pages, then answers only from them. Nothing is stored after the demo.

Built by us, running on this page

You're already talking to one.

The agent at the bottom of this page is an aumiqx build, on the same pattern we ship for clients. Open it and push it: ask about prices, give it a domain to audit, talk to it, or ask for a person.

  • Answers from our facts only

    It knows our services and from-prices and nothing more. Ask it something off-script and it says it doesn't know.

  • Shows the next step, not just text

    Service cards, a booking slot, a lead form or a live audit appear inside the conversation, not behind a link.

  • Does real work on the spot

    Give it any domain and six agents pull rankings, competitors and AI visibility, with the report ready in about a minute.

  • Listens

    Tap the mic and talk. It sends when you stop. The audio is turned into text on the fly and never stored.

  • Knows when to stop

    Ask for a person, a contract or a custom price and it books you with Axit instead of guessing. Details are only saved with a ticked consent box.

  • Costs cents, not dollars

    Free models first, a paid model only as a fallback, and every answer checked for shape before you see it.

The short answer

An AI chatbot development company builds an assistant that answers questions from your own content, on the channels your customers or staff already use, and passes anything risky to a person. Platforms cost $19 to $500 a month; agencies that publish prices start around $1,718; cost guides quote $20,000 to $80,000 for a custom build.

Published entry price
$1,718
median of 7 agencies
Custom bot on your docs
$20k–80k
most cost guides
Model cost, 1,000 chats
$2–50
openrouter, Sep 24, 2026
Voice, all-in
$0.10–0.25
per minute, typical

Definition

What does an AI chatbot development company do?

It connects a language model to the content your business already has, so the bot searches that content for every question and answers from what it finds. This is called retrieval-augmented generation, or RAG. Most bad chatbots aren't bad because of the model. They search the wrong documents, or old ones, or they answer confidently when nothing relevant was found. So most of our work goes into the search step, the test set and the handoff, not the prompt.

What we build

  • Knowledge pipeline. Your docs, help centre, PDFs and product data cleaned, split and indexed for retrieval, with a way to keep it in sync when they change.
  • The assistant. Answers with citations, a tone that matches yours, clear refusals outside its scope, and a handoff to a human with the conversation attached.
  • Channels. A website widget, WhatsApp, Slack or an API your own app calls. Same brain behind each one.
  • Quality review. A test set of real questions, a log of unanswered ones, and a simple way to see what people ask that your docs do not cover.

What you own at the end

  • The code, prompts and test set, in your repositories.
  • The search index and conversation logs, in your cloud and your name.
  • The model accounts, so you see every bill directly.
  • A runbook for updating sources, and the option to keep us on a retainer, or not.

Honest comparison

Chatbot, AI agent or voice agent: which do you need?

If people need information, build a chatbot. If they need something done, like a refund issued or a booking moved, they need an AI agent with access to those systems and a person approving risky actions. If your customers phone rather than type, it's a voice agent. Many projects start as a chatbot and add a few narrow actions once the answers are trusted.

TypeWhat it doesWhere it runsPriced perUse it when
AI chatbotAnswers questions from your contentWebsite, WhatsApp, Slack, TeamsPer conversationPeople need information
AI agentTakes actions in your systemsAnywhere, often behind a chatbotPer step, in tokensPeople need something done
AI voice agentAnswers, qualifies, books and transfers callsA phone numberPer minuteCustomers call instead of typing

Buy side · prices fetched Sep 24, 2026

Buy a chatbot platform or build a custom one?

Plenty of chatbots should be bought. If your answers live in a help centre and your helpdesk vendor sells a bot, start there. These are the list prices on each vendor's own page today.

PlatformPriced perPrice
Intercom Fin per resolved conversation$0.99 per outcome, plus seats from $29/mo
Zendesk AI agents per automated resolution$1.50 committed, $2.00 pay as you go
Tidio (Lyro) per plan and AI conversation volumeFrom $24.17/mo; Plus from $300/mo
Chatbase per message credit$40 to $500/mo (700 to 15,000 credits)
Botpress per conversation$150/mo for 250, then $0.50 to $0.65 each

The volume rule (our model, maths shown)

At 3,000 resolved conversations a month, Intercom Fin's $0.99 per outcome is $2,970 a month, or $35,640 a year, and Zendesk's $1.50 is $54,000 a year. A custom build at the top of the agency range ($15,000) plus $1,000 a month to run is $27,000 in year one and $12,000 after that. Under about 1,000 a month, the fee is under $1,000 a month and the platform's inbox, analytics and updates are worth it.

When a platform bot is enough

Your content is public and tidy, you need one channel, and the handoff is to the same vendor's inbox. Our guides to customer service AI agents and free chatbots worth using are a better next step than a call with us. Build when the sources are private or messy, the bot spans several channels, or the answers must be tested.

Sourced · dated Sep 24, 2026

What AI chatbot development costs in 2026

There are three ways to get a chatbot, and they differ by a factor of a hundred. Most of the difference is scope, not skill.

PathWhoBuild costTime to launchRunning cost
Chatbot platform (SaaS)Tidio, Chatbase, Intercom Fin, Botpress$0 to set upA day to 2 weeks$19 to $500/mo, or $0.99 to $2 per resolution
Agency build on your docsSmaller agencies, including usAbout $1k to $15k2 to 6 weeks$100 to $1,000/mo
Custom build by a dev shopSoftware firms, enterprise scope$20k to $150k and up2 to 6 monthsPlus 15 to 20% of the build per year

Our starting price · AI chatbot

From $2,500

A chatbot that answers from your docs, cites its sources and hands off to a person. Scope sets the final quote: more sources, more channels, integrations and languages move it up.

Updated 2026-09-24

What agencies publish vs what cost guides quote

7 agencies in our 2026 survey publish a chatbot price. Their entry prices have a median of $1,718, and the middle half start between $1,360 and $4,403. Most cost guides quote $20,000 to $80,000 for a chatbot on your own documents. Both are right: guides describe enterprise scope with several integrations and months of work, while published price lists describe a defined bot on a known stack. Ask any quote which one you are buying.

$1k$10k$100k

Agencies' published entry prices (n=7)

$1k–$4k

Topflight

$35k–$80k

Cleveroad

$18k–$80k

WotNot

$20k–$80k

Appinventiv

$80k–$200k

Denser

$2k–$30k

Quickchat

$75k–$500k
Lavender: the middle half of chatbot entry prices that 7 agencies publish (median $1,718). Grey: what each cost guide quotes for a chatbot on your own knowledge base. Log scale. Guides fetched 2026-09-24.
The 7 published chatbot prices

Cost per 1,000 conversations at today's model prices

Assumes a four-turn support conversation of about 20,000 input tokens (your retrieved content, sent each turn) and 1,000 output tokens, at OpenRouter list prices on Sep 24, 2026. Per-resolution platforms are shown for scale. Hosting, monitoring and keeping the index current usually add $100 to $1,000 a month; see our API pricing guide for how tokens are billed.

Model or platformInput / output, per 1M tokensPer 1,000 conversations
DeepSeek V4 Flash$0.089 / $0.177$1.96
GPT-5.4 Nano$0.2 / $1.25$5.25
Gemini 3.1 Flash Lite$0.25 / $1.5$6.50
GPT-5.4 Mini$0.75 / $4.5$19.50
Claude Haiku 4.5$1 / $5$25.00
Claude Sonnet 5$2 / $10$50.00
Intercom Fin$0.99 per outcome$990
Zendesk AI agents$1.50 to $2.00 per resolution$1,500

What makes a chatbot cost more

  • Sources. Many formats, contradictions or content behind logins take longer to clean and index.
  • Integrations. Handoff to a helpdesk, CRM lookups or order status add build and test time.
  • Channels. WhatsApp needs Meta-approved templates outside the 24-hour window; voice needs telephony.
  • Languages. Each language needs its own test questions, even when the model speaks it.
  • Compliance. Data residency, retention rules and audit logs for regulated industries.
  • Permissions. Internal bots must respect who can see which documents.

By use case

AI chatbots we build, by use case

01

Customer service chatbots

Answer order, account and policy questions from your help centre, and pass the rest to Zendesk, Intercom or Help Scout with the conversation attached.

Hands off to a person: Refunds, complaints and anything account-specific.

02

Website chatbots for leads and bookings

Answer pre-sales questions from your site and pricing pages, qualify the visitor and offer a booking link.

Hands off to a person: Custom quotes and anything not published.

03

Ecommerce chatbots

Order status, shipping, returns and product questions, from your store data and policies.

Hands off to a person: Damaged orders and exceptions to the returns policy.

04

Internal knowledge base chatbots

Answer staff questions in Slack or Teams from your wiki, docs and tickets, with a citation for every claim and access that follows each person's permissions.

Hands off to a person: HR and legal questions go to a named owner.

05

WhatsApp chatbots

The same knowledge on the WhatsApp Business API. Replies are free-form for 24 hours after a customer writes; outside that window only Meta-approved templates can be sent.

Hands off to a person: Anything needing a person, inside the 24-hour window.

Priced per minute · fetched Sep 24, 2026

AI voice agents and AI receptionists

A voice agent answers your phone number, answers questions from the same knowledge base, qualifies the caller, books a slot and transfers to a person with a summary. Voice breaks in places text doesn't: half a second of delay feels rude, callers interrupt, accents and background noise confuse transcription, and a botched transfer loses the customer. Those four are what we test before launch.

VendorPriceNotes
Vapi $0.05/min platformPlus speech, model and phone line; their 1,000-minute example totals $82 to $129/mo
Retell AI $0.07 to $0.31/minDepends on the model and voice; their worked example is $0.23/min
Bland $0.12 to $0.14/min$0.12 on the $299/mo plan, $0.14 with no platform fee
ElevenLabs agents $0.08/minSpeech engine for agents
Intercom Fin Voice $1.99 per outcomeSold through Intercom sales
Smith.ai AI Receptionist $1.67 to $3.00 per callDone for you, priced per call
Synthflow From $30,000/yrEnterprise contracts only; no self-serve prices shown

Platform prices look small per minute, but the platform fee is only one layer: speech-to-text, the model, the voice and the phone line are each billed on top, which is why a typical all-in cost lands around $0.10 to $0.25 a minute. Buy a done-for-you receptionist if you need call answering this week. Build when the agent must read your own calendar, CRM or case system, or when minutes are high enough that per-call fees add up. Comparing voices? Our ElevenLabs review covers the speech layer.

AI receptionists for clinics, law firms and home services

These three lose the most to missed calls. A clinic receptionist books and reschedules without touching clinical questions. A law firm intake agent collects the facts of a matter and runs a conflict check list, but never gives advice. A home services agent quotes call-out fees you publish and books the visit. In each, anything outside the script goes to a person.

Our starting price · AI voice agent

From $6,000

A voice agent on a phone number, with call summaries and a transfer to a person. Scope sets the final quote: call volume, integrations, languages and after-hours rules move it.

Updated 2026-09-24

Anatomy of a build

How we build a chatbot, in five steps

  1. 01

    Audit the sources

    We read what the bot will answer from, and flag gaps and contradictions before building anything.

  2. 02

    Build retrieval

    Indexing and search tuned on your real questions, because retrieval quality decides answer quality.

  3. 03

    Write the test set and the boundaries

    Real questions with the right answers, plus what it must refuse and when it hands off to a person.

  4. 04

    Launch narrow

    One channel and one audience first, with every unanswered question logged.

  5. 05

    Widen from the logs

    New sources and channels are added where the logs show people asking and the bot not knowing.

From you it needs access to the content it should answer from, someone who can say whether an answer is right, and a person or queue to receive handoffs. Where several assistants need the same systems, we expose them once as an MCP server.

Hallucination control

How we stop it making things up

In 2024 a Canadian tribunal held Air Canada liable after its website chatbot told a customer he could claim a bereavement fare refund after travel, which the airline's actual policy didn't allow. The airline argued the chatbot was responsible for its own words. It lost. Whatever your bot says, you said.

Answers only from retrieved sources

Each answer is built from passages found in your content for that question, not from what the model remembers.

A citation on every answer

The page or document it used is shown, so a person can check it in one click.

An honest 'I don't know'

When nothing relevant is found, it says so instead of guessing. You saw that in the demo above.

A handoff with context

Refunds, complaints and personal data go to a person, with the conversation attached.

A test set run on every change

Real questions with known answers, re-run whenever the prompt, the model or your content changes.

Disclosure and liability

Yes, with three things to get right. Every bot we build says it is a bot in its first message, and conversation logs are treated as personal data under UK GDPR and EU GDPR wherever they contain it: kept only as long as needed and deletable on request.

Say it's a bot (California)

California's BOT Act makes it unlawful to use a bot to mislead a person in California about its artificial identity in order to incentivise a purchase or influence a vote. A clear, conspicuous disclosure that it is a bot removes the liability.

Cal. Bus. & Prof. Code 17940 to 17943 (SB 1001, 2018)

Say it's AI (European Union)

Article 50 of the EU AI Act requires AI systems that interact directly with people to be designed so those people are informed they are talking to an AI, unless that is obvious. The Act applies from 2 August 2026.

EU AI Act, Article 50(1) and Article 113

You own what the bot says

In Moffatt v. Air Canada, a British Columbia tribunal held Air Canada liable for negligent misrepresentation after its website chatbot gave a customer wrong information about bereavement fares, and rejected the argument that the chatbot was a separate entity responsible for its own actions.

Moffatt v. Air Canada, 2024 BCCRT 149

This is a summary, not legal advice. Check the rules where you operate with a lawyer before launch.

Interactive checklist

10 questions to ask any chatbot or voice agent developer (including us)

Tick the ones they answered well. If they can't show you a live bot, or tell you what it costs to run, keep looking.

Tick each question the agency answered well.

0/10

Before you call

Good fit, and not a fit

  • Customers or staff ask the same questions and the answers already exist in your docs
  • Support is buried in repetitive questions that stop people getting to the hard ones
  • You need answers grounded in your content, not in whatever the model remembers
  • Your documentation is thin or out of date. The bot can only be as good as its sources
  • Every question needs account-specific actions. That is an agent, not a chatbot
  • You want the bot to replace your support team entirely on day one

Questions

Frequently asked

How much does it cost to build a custom AI chatbot?+

It depends on the path. Chatbot platforms cost $19 to $500 a month or about $1 to $2 per resolved conversation. Agencies that publish prices start at a median of $1,718 (7 agencies, 2026), and cost guides quote $20,000 to $80,000 for a custom chatbot on your own documents. Our chatbots start from $2,500 and voice agents from $6,000; scope sets the final quote.

How much does an AI chatbot cost to run each month?+

The model is the cheap part: at current prices a four-turn support conversation costs about $2 to $50 per 1,000 conversations, depending on the model. Hosting, monitoring and keeping the index current usually add $100 to $1,000 a month. Per-resolution platforms like Intercom Fin charge $0.99 per resolved conversation instead.

How long does it take to build an AI chatbot?+

A platform bot can go live in a day. An agency build on your own documents usually takes 2 to 6 weeks, and a custom enterprise build 2 to 6 months. Most of the time goes into cleaning sources, tuning retrieval and testing on real questions, not the model.

Can I build my own AI chatbot instead of hiring someone?+

Yes. Tools like Chatbase, Tidio and Botpress let you point a bot at your site in an afternoon. Hiring pays off when the bot must answer from messy or private sources, hand off properly to your helpdesk, run on several channels, or be tested and maintained.

Should I use Intercom Fin, Tidio or Chatbase, or a custom chatbot?+

Our rule of thumb: under about 1,000 resolved conversations a month, buy a platform. Above about 3,000, per-resolution fees usually cost more within a year than a custom build plus its running costs. In between, it depends on integrations, data sensitivity and how much control you need over answers.

Can a custom chatbot be trained on our own data?+

Yes, though usually not by retraining the model. Your documents are indexed so the chatbot searches them for each question and answers from what it finds, with the source cited. It is cheaper than fine-tuning and stays current when your docs change.

What stops the chatbot from making things up?+

It answers only from passages retrieved for that question, shows its sources, says it doesn't know when nothing relevant was found, and hands off to a person for anything risky. A test set of real questions is re-run whenever the prompt, the model or your content changes.

Are AI chatbots legal to use with customers?+

Yes, with conditions. California's BOT Act requires a clear disclosure when a bot is used to encourage a sale, the EU AI Act requires people to be told they are talking to an AI from 2 August 2026, and a Canadian tribunal held Air Canada liable for its chatbot's wrong answer. This is not legal advice; check the rules where you operate.

Is ChatGPT a chatbot or an AI agent, and which do I need?+

ChatGPT is mainly a chatbot with some agent features. If your customers need information, you need a chatbot that answers from your content. If they need something done, like a refund issued or a booking moved, you need an agent with access to those systems and a person approving risky actions.

Can it hand off to a human, and where can it run?+

Yes. It passes the conversation, with context, to your helpdesk, inbox or a Slack channel. It can run on your website, WhatsApp, Slack or Teams, or behind an API your product calls, with one knowledge base behind all of them.

How much does an AI voice agent or AI receptionist cost?+

Voice platforms charge about $0.05 to $0.31 per minute, and a typical all-in cost is roughly $0.10 to $0.25 per minute. Done-for-you receptionist services charge $1.67 to $3.00 per call. Our voice agents start from $6,000, on a phone number with call summaries and a transfer to a person.

Who owns the chatbot, the data and the conversation logs?+

You do. The code, prompts, index, test set and logs live in your repositories and accounts from the first day.

Bring your docs. We'll tell you if a platform will do.

A 30-minute call with the person who would build it. You leave knowing whether to buy, build or wait, and roughly what it would cost to build and to run.

Not sure yet? Start with consulting Building a product?