The short answer
An AI agent is software that understands what a customer is asking, decides what to do, and then takes action: it answers the question, books the appointment, updates your records or hands the conversation to a person. Unlike a simple chatbot that follows a script, an agent works towards a goal using your real information, around the clock.
Key takeaways
- An AI agent understands requests, decides the next step and takes action, not just replies.
- It works from your approved information and tools, and hands sensitive cases to a person.
- At PJW Partners, agents start from $1,000 USD for one agent on one channel.
- Most single agents go live in 2 to 3 weeks, after training and testing on real questions.
How does an AI agent actually work?
An AI agent works by combining a language model, your business knowledge and a set of tools it is allowed to use. Every agent has these three parts working together:
- A language model that understands messages written the way people really write, typos and all, in English, Thai and other languages.
- Your knowledge: services, prices, opening hours, policies and the tone your business uses. The agent answers from this, not from guesswork.
- Tools it is allowed to use, such as your booking calendar, your CRM, a spreadsheet or your inbox. This is what lets it act instead of only replying.
When a message arrives, the agent works out what the person wants, checks your information, chooses the right next step and carries it out. Anything sensitive, unusual or high-value goes to a person on your team.
The part that makes it an agent is the goal. A plain AI assistant answers and stops. An agent is given an outcome to work towards, such as “get this enquiry booked” or “collect the details our lawyers need”, and keeps going, one sensible step at a time, until it gets there or hands over.
What happens when a customer messages an AI agent?
A customer message sets off a short loop: the agent reads it, checks your information, asks what it still needs, acts and then reports back. Here is a typical example, a customer writing to a beauty clinic on LINE at 10 pm asking about a facial and whether Saturday is free:
- It reads the message and works out the language and the intent: a question about a treatment, plus a booking request.
- It checks your approved information for that treatment: what it involves, how long it takes and the price you have published.
- It asks for what is missing, such as a preferred time on Saturday, instead of guessing.
- It checks your calendar through a connected tool and offers the slots that are genuinely free.
- It books the appointment and sends a confirmation, with a reminder set for the day before.
- It logs the enquiry in your CRM or spreadsheet and lets your team know a new booking has arrived.
- It hands over if the customer asks something it should not answer, such as a medical question, passing your team a short summary so nobody has to start again.
The customer gets an answer and a booking in a few minutes, at a time when nobody on your team is working.
What can an AI agent do for a small business?
An AI agent can take on almost any task that repeats every day and follows clear rules. The most useful jobs for small and growing businesses are:
- Answering enquiries on your website, LINE, WhatsApp, Instagram or Facebook, in seconds and at any hour.
- Booking appointments straight into your calendar, with confirmations and reminders.
- Qualifying leads: asking the right questions and passing the serious ones to your sales team with a summary.
- Following up with quotes, reminders and review requests, so nothing slips.
- Admin such as logging enquiries in a spreadsheet, sending invoices or preparing simple reports.
Clinics use agents to book consultations after hours. Gyms use them to answer questions about classes and prices. Professional firms use them to collect the details of a new case before the first call.
What an agent should not do is just as important. It should not give medical, legal or financial advice, agree to discounts you have not approved, or make promises your team cannot keep. Those limits are written into the agent before it ever speaks to a customer.
AI agent, chatbot or AI assistant: what is the difference?
A chatbot follows a script, an AI assistant understands and answers, and an AI agent understands, decides and acts. The words get mixed up, so here is the short version:
| Type | What it does |
|---|---|
| Rule-based chatbot | Follows a fixed script or button menu. Breaks when a question doesn’t fit. |
| AI assistant | Understands free-text questions and answers them, but usually can’t take action. |
| AI agent | Understands, decides and acts: books, updates, follows up and escalates to a person. |
We go deeper in AI agent vs chatbot: which does your business need?
Is an AI agent safe to use with customers?
Yes, an AI agent is safe to use with customers when it is built with clear limits, tested properly and reviewed after launch. A well-built agent is designed with those limits from day one:
- It answers only from information you have approved, and says so when it doesn’t know.
- It hands anything sensitive, such as complaints, medical or legal questions and refunds, to a person.
- It only touches the systems it needs, and keeps personal data within privacy laws such as Thailand’s PDPA and the GDPR.
- Every conversation can be reviewed, so you can see what it said and improve it.
It should also be honest about what it is. We recommend telling customers they are talking to an AI and always offering a quick route to a person. If your agent will handle personal details, read AI agents and the PDPA before you launch.
What do you need to prepare before launching an AI agent?
You need your business information, a clear goal and a short list of rules for when the agent should hand over. Most of this already exists in your team’s heads, inboxes and price lists; the work is writing it down in one place. Before we build, we ask clients to gather:
- The questions customers ask most, in their own words. Old chats and emails are the best source.
- Services, prices and timings exactly as you want them quoted, including what is and isn’t included.
- Policies: cancellations, deposits, opening hours, locations and anything else customers check before they commit.
- The goal for the agent, such as booked consultations, qualified leads or fewer repeat questions.
- Handover rules: which topics always go to a person, and who on your team receives them.
- Access to your tools, such as your calendar, booking system or CRM, so the agent can act.
- Your tone: a few examples of replies you like, so the agent sounds like your business.
Our guide to training an AI agent on your business walks through each step in detail.
How much does an AI agent cost, and do I need a website first?
At PJW Partners, AI agents start from $1,000 USD for one agent on one channel, such as your website chat or LINE. Most single agents go live in 2 to 3 weeks: we learn how your business works, build and train the agent on your information, test it with real questions, and launch it with you watching.
Bigger systems, with several agents working across channels and tools, get a tailored quote.
You don’t strictly need a website first, but the two work best together. Your website wins the customer’s trust; the agent answers them the moment they have a question. Most of our clients launch a website first and add an agent once visitors start coming in. If you are planning both, read how much a website costs.
Not sure which tasks to hand to an agent first? Tell us about your business and we will reply within one business day with a suggestion.
Questions
Will an AI agent replace my staff? +
No. A good agent takes the repetitive work, such as the same questions, bookings and reminders, so your team can spend their time on customers who need a person.
Can an AI agent speak Thai and English? +
Yes. Modern agents understand and reply in both, and switch to the customer’s language automatically.
Which channels can an AI agent work on? +
Your website chat, LINE, WhatsApp, Instagram, Facebook Messenger and email, and with the right setup, the phone.
What happens when the agent doesn’t know the answer? +
It says so honestly and passes the conversation to your team with a summary, instead of guessing.
Do I need technical skills to run an AI agent? +
No. Once it is live, your main job is to tell us when prices, services or policies change and to read the conversations it hands over.
Can I change what the agent says after launch? +
Yes. The agent answers from information you approve, so updating that information changes its answers without rebuilding anything.
Work with PJW Partners
Websites and AI agents that bring in customers.
Websites from $1,500 and AI agents from $1,000. Tell us about your business and we’ll reply within one business day.