AI customer assistant: a practical guide.

Modern AI assistants understand meaning and context — they aren't the rule-based chatbots of 2018. Here's what they can and can't do, which channels matter, how to deploy one, and how to think about cost.

9 min read No-code

AI assistant = chatbot?

Old chatbots (2015-2020) were simple rule-based systems: "if this word appears, return this answer." A customer typed "hours" and the system returned a time table; if they typed "when are you open" the bot froze.

Modern AI assistants — built on large language models like GPT, Claude, and Gemini — understand meaning and context. "Are you open tomorrow at 3?" → "Yes, tomorrow is Wednesday and we're open 9 AM to 7 PM."

That's a huge leap, but AI still can't do everything.

What AI can do

1. Answer frequently asked questions. "What are your hours?", "What's your address?", "Is there parking?", "What's your return policy?" — these are typically 60-70% of inbound traffic.

2. Book appointments. "I'd like an appointment for tomorrow at 2 PM" → AI checks the calendar, confirms availability, books it.

3. Product recommendations. "Which moisturizer do you recommend for oily skin?" → looks at the catalog, filters, recommends.

4. Order tracking. "My order number is 12345, where is it?" → pulls the status from the API.

5. Form filling. "I'd like to sign up" → AI collects information, pushes it to the system.

6. Translation. "What does this contract say?" → summarizes, translates.

What AI can't do

1. Empathy-heavy situations. "I've wasted time, I'm angry" — the AI's response feels artificial. Hand off to a human.

2. Complex complaint resolution. A product broke three times and the customer wants compensation — the AI doesn't have authority.

3. Unusual requests. "Dear manager, my son is getting married — could you arrange a custom menu for 200 people, I'll pay an extra 30%?" — AI isn't designed for that level of nuance.

4. Sensitive topics. Medical advice, legal counsel, financial planning — if the AI is wrong, the company is liable.

5. Creative work. Building a personal relationship with a customer, closing a deal — that's still human work.

Which industries benefit most?

High benefit:

Low benefit:

Channels

Website: Bottom-right chat bubble. The most common.

WhatsApp: 96% adoption in markets like Turkey, India, and Brazil. AI connects via the WhatsApp Business API.

Instagram DM: Strong for reaching younger audiences.

Voice phone: Strong in English, weaker in many other languages — but improving fast.

In-app chat: A chat window inside your native mobile app.

Ideal: one AI managing every channel. A customer starts on WhatsApp, continues on the web, and the AI remembers the thread.

Handoff strategy

The hardest moment for an AI: knowing its own limits and calling for a human.

A good handoff:

Customer: "The product arrived 3 months ago and is still broken. I've gone through every step with the company. I'm taking this to consumer arbitration. Consider this message your final warning."
AI: "I understand how exhausting this has been. I'm escalating you to a customer relations specialist who will reach out within 5 minutes."

The worst-case response from an AI is something like "Could you rephrase that more politely?"

Deployment timeline

  1. Data preparation (1 week): Compile your FAQ documents, product descriptions, policies.
  2. AI selection (1 day): OpenAI's GPT-4? Anthropic's Claude? Open-source models? Decide based on cost and security factors.
  3. Training (1-2 weeks): Fine-tune the model with your data. Measure responses against test questions.
  4. Pilot (2 weeks): Web only, limited hours. Review the logs.
  5. Gradual rollout (1 month): Web first, then WhatsApp, then Instagram.
  6. Continuous improvement: Analyze conversation logs monthly, update the training.

Cost breakdown

Annual cost of an AI customer assistant typically includes:

First-year total: $2,000-6,000. In exchange: not hiring a half-time support agent (~$25,000-40,000/year) plus 24/7 availability.

The AI assistant in the Onremo Store

AI Customer Assistant (by Helio AI) is the most popular AI product in the Onremo Store:

$39/month. Up to 5,000 conversations included per month; $0.02 per conversation above that.

Next step

The right time for an AI customer assistant: 30+ daily support requests and people asking the same 5-10 questions over and over.

To browse AI systems in the Store, go to the Store and filter by the "AI" category.

Related: how to buy a system from the Store.

Start your own story on Onremo.

Create an account, explore the Store, get one-click access to every system.

Go to Onremo →