ISO 9001 · ISO 27001 · ISO 20000 certified CERT-In aligned security audits

ETHICSCOMPUTER / JOURNAL · ARTIFICIAL INTELLIGENCE & AUTOMATION · 14 NOV 2026 · 5 MIN READ

How to Choose an AI Chatbot Development Company: Questions to Ask

Every chatbot vendor has a good demo. These questions about accuracy, data, integrations, ownership and running costs show you which proposal will survive real customers.

IN THIS ARTICLE (9 Sections)

Almost every chatbot vendor can show a good demo. The demo answers smoothly because it was built on a handful of clean documents and rehearsed questions. Your customers will not be so kind. They will ask in Hinglish, paste order numbers with typos, ask about a policy that changed last week and expect a person when the bot cannot help.

The questions you ask before signing decide whether you get a bot that handles that reality or one that is switched off after a month. This guide lists what to ask an AI chatbot development company, what a good answer sounds like and the red flags that should make you walk away. It applies whether you want a website bot, a WhatsApp assistant or an internal HR helpdesk.

Quick answer: Judge an AI chatbot vendor on six things: how they test accuracy on your real questions, how the bot handles questions it cannot answer, how your data is protected, which systems it will integrate with, who owns the code and content, and what it will cost to run each month. Ask for a pilot on your own data with an agreed accuracy target before you commit to a full build.

Questions about accuracy and testing

  • Will you test the bot on our real customer questions before launch? A good vendor asks for chat logs, emails or ticket exports and builds a test set from them, not from questions they wrote themselves.
  • What accuracy target will we agree on, and how will you measure it? Look for a clear method: a question bank, a definition of a correct answer and a pass mark agreed in writing.
  • What does the bot say when it does not know? The right answer is that it admits it and offers a person, rather than guessing.
  • How do you handle Hindi, Hinglish and spelling mistakes? Ask to see examples during the demo, not just a yes.

Questions about data and privacy

  • Where will our documents and conversations be stored, and for how long? You should get a specific answer, including the cloud region and the retention period.
  • Which AI provider will process the messages, and does it use our data for training? Business API tiers of the major providers state that API data is not used for training by default. Ask the vendor to show you the setting and the terms for your plan.
  • How is personal data in chats handled? Phone numbers, addresses and order details are personal data under the Digital Personal Data Protection Act, 2023. Ask about masking, retention and who on the vendor's team can read transcripts.
  • Will you sign an NDA and a data processing agreement? A serious vendor will not hesitate.

Questions about integrations and handoff

  • Which of our systems will the bot read from and write to? Order status, appointment booking and lead capture all need integration with your CRM, helpdesk, calendar or order database. A bot that can only answer FAQs has limited value.
  • How does handoff to a person work? Your agent should receive the full conversation, the customer details and the reason for escalation, on whichever channel your team uses.
  • Can the same assistant run on our website, app and WhatsApp? Separate bots per channel drift apart and start giving different answers.
  • Can our team update answers without a developer? Ask to see the admin panel itself.

Questions about ownership, cost and support

  • Who owns the source code, prompts and knowledge base? If the bot lives only inside the vendor's platform, moving away later can mean starting again.
  • What will it cost to run each month, and what drives that cost? Running cost depends on conversation volume, message length, the model used and channel charges such as WhatsApp Business Platform messaging fees. Ask for an estimate based on your own volume.
  • What happens after launch? Monitoring, a monthly review of unanswered questions and content updates should be part of the plan, not an afterthought.

Compare proposals side by side

Put the proposals you receive into a simple comparison like this one. Gaps become obvious quickly.

What to checkGood signRed flag
Accuracy testingTest set built from your real queries, pass mark in writingA headline accuracy figure with no test on your data
Unknown questionsBot admits it and offers a personBot always produces some answer
Data handlingNamed AI provider, retention settings, NDA offeredVague answers about where data goes
IntegrationsSpecific systems, APIs and fields listedIntegrates with everything, no detail
OwnershipCode, prompts and content handed overLocked into a platform with no export
Running costMonthly estimate based on your volumeOnly a build price, nothing about usage
After launchMonthly tuning from missed questionsSupport only for bugs

Red flags that should end the conversation

  • Promises that the bot will never give a wrong answer. Honest vendors explain how wrong answers are caught and routed instead.
  • A fixed price for a complex, integrated bot quoted without asking to see any of your data.
  • Refusal to run a small pilot or proof of concept first.
  • Pressure to load all customer data into a tool before any privacy discussion.

If you are still deciding between a chatbot that answers and an agent that takes actions, our article on autonomous AI agents for business explains the difference. For the build itself, see our AI chatbot development service.

How EthicsComputer can help

Our AI chatbot development proposals answer each of these questions in writing. We start with a pilot on your real queries, agree the accuracy target with you and hand over the source code and knowledge base. Knowledge-based chatbots typically go live in three to five weeks, and deeply integrated assistants in six to ten weeks. If WhatsApp is your main channel, see WhatsApp AI automation, and for phone calls, AI voice bot development.

Comparing chatbot proposals right now?

Send us your requirement and the questions your customers ask most, and we will explain how we would handle accuracy, data and integrations for your case.

About the author

Vinay Grover founded EthicsComputer in Rohtak, Haryana in 2018 and leads its cybersecurity, software and AI work for clients across India and abroad. EthicsComputer is ISO 9001, ISO/IEC 27001 and ISO 20000-1 certified.

Vinay Grover
AUTHOR

Founder & CEO of EthicsComputer (Rohtak, since 2018). Leads the company's VAPT, software architecture and AI automation work. Full profile

END OF ARTICLE

Continue exploring EthicsComputer's technical journal.

EthicsComputer assistant
EthicsComputer Assistant
Automated assistant (AI) • Team replies in office hours
⚡ Fast Response
Instant Quote & Demo
Get a Quote
Verified by SMS
Call Direct
+91 90532 10052
Direct Architect Scoping // Step 1 of 2

Request Fast Quote & Architecture SLA

Receive a preliminary project architecture, pricing tiers, and timeline estimate within 24 hours under strict NDA.

100% Mutual NDA Protected Step 1 of 2 (3 fields)

We use your phone number, company and the details you choose only to reply to your enquiry and prepare a quote. We do not sell your data. Privacy Policy · Your rights