Conversational AI for a small business works best when it starts narrow: one channel, such as WhatsApp or website chat, and a short list of repetitive questions about hours, prices, bookings and delivery. Modern chatbots built on large language models can answer those well if they read from your own approved information and hand over to a person when unsure. Start there before automating anything that touches payments or complaints.
What is conversational AI in customer service?
Older chatbots followed decision trees: press 1 for opening hours, press 2 for prices. Conversational AI uses language models to understand free-form questions and reply in natural sentences. The better systems also look up information in your documents or booking system before answering, so replies are grounded in your facts.
Typical tools for smaller teams include the AI features in help desks such as Intercom Fin, Zendesk AI and Freshdesk Freddy, website chat tools like Tidio and Crisp, WhatsApp Business platforms with AI add-ons, and custom setups built with OpenAI, Anthropic or Google models. Features and pricing vary widely, so compare current plans.
Which questions should a small business automate first?
Look at your last month of messages and sort them. Most small businesses find that a handful of question types make up a large share of the volume.
- Opening hours, holiday closures and location directions.
- Prices, packages and what is included.
- Booking, rescheduling and cancellation rules.
- Delivery times, areas and costs.
- Order status, if your platform exposes it safely.
- Simple product questions such as sizes, ingredients or compatibility.
Leave out complaints, refunds, medical or legal questions and anything emotional. Those should go straight to a person, and the bot’s job is to collect details and route them quickly.
A small rule of thumb helps here: if a question has one correct answer that does not depend on the customer’s personal situation, it is a good candidate. If the answer starts with “it depends”, keep a person involved.
Which channel should you start with?
Start where your customers already message you. For many local businesses in Türkiye and across Europe, that is WhatsApp. For e-commerce stores, website chat and email often matter more. Instagram direct messages are important for beauty, fashion and hospitality.
WhatsApp automation runs through the WhatsApp Business Platform via Meta or an approved provider, with rules on message templates and opt-in. Website chat is simpler to set up and test. Pick one, get it right, then add the next.
How do you set it up without a developer?
- Write a knowledge base of 20 to 40 short articles: one question, one clear answer, with prices and conditions in text.
- Choose a tool that can answer only from that knowledge base, and turn on that restriction.
- Write a hand-over rule: when the bot is unsure, or the customer asks for a person, it says so and creates a ticket or notifies staff.
- Set business hours behaviour, so customers know when a human will reply.
- Test with 30 real questions from past messages, including awkward and misspelled ones.
- Launch to a portion of traffic or one channel and review conversations daily for the first two weeks.
Tone and language
Give the assistant a short style note: friendly, brief, no slang, use the customer’s language. If you serve both Turkish and English speakers, test both. Models handle many languages well, but your knowledge base should still include local terms customers actually use.
What does it cost, and how do you judge it?
Costs come from the tool subscription, per-conversation or per-resolution fees on some platforms, messaging fees on WhatsApp, and your time to write and maintain the knowledge base. Check the pricing model carefully: per-resolution pricing is predictable for low volume but can rise quickly in busy seasons.
Judge it by outcomes you can count. Track the share of conversations resolved without staff, response time outside working hours, bookings or orders started from chat, and customer satisfaction ratings if the tool offers them. Compare with the staff time those questions used to take.
What guardrails keep it trustworthy?
- The bot states that it is an automated assistant.
- It never invents prices, availability or policies; if the answer is not in the knowledge base, it hands over.
- Personal data is collected only when needed and handled under KVKK or GDPR rules.
- Staff review a sample of conversations every week and correct the knowledge base.
- There is always a visible way to reach a human.