Chatbots are everywhere on the web now. Most likely, just about any company you visit that has some size to it, they will have a chatbot.
They are usually in the bottom right corner of websites you visit, and it will be a little bubble that says something like “Hi there! How can I help you today?”
Since they seem to be everywhere, you’re probably wondering
Should I have one of those on my site?
The honest answer is: it depends.
Done well, a chatbot can save you real time, capture leads around the clock, give your visitors instant answers, and even encourage them to purchase or take that next step in your sales funnel, even at 2 a.m. on a Saturday.
Done poorly, it can frustrate your visitors, confuse them, and make your brand look bad in the process, which is something many people don’t really consider.
So let's walk through when a chatbot actually helps, when it doesn't, and what you need to do to set one up the right way.
When a Chatbot Actually Helps
The best use cases for chatbots are simple and repetitive.
Think about the questions your team answers over and over. Things like:
- What are your hours?
- Do you ship to Canada?
- What's your return policy?
- How do I request a [fill in the blank]?
- What products do you carry?
- How do I [fill in the blank]?
- How do I get in contact with someone?
If you're answering those questions on a regular basis (it doesn’t have to be multiple times a day, but at least weekly) either by phone, email, or live chat, or you want to appear as a bigger company than you are, a well-trained chatbot can handle that load for you. And it can do it 24/7 without getting tired or taking lunch.
E-commerce sites are a great fit. So are businesses with a clear set of FAQs: service companies, manufacturers, specialty product sellers, and anyone with a well-defined niche.
We recently set one up for a client in the industrial tarp industry. Their company gets questions constantly about product specs, compatibility, repair requests, and pricing. A chatbot was a natural fit because the questions were predictable and the answers were clear.
That's the sweet spot.
When a Chatbot Is NOT the Right Call
However, while it may sound great, a chatbot is not for everyone.
If your business is highly consultative, where every client situation is unique and requires a real conversation, a chatbot probably isn't going to help you much. It may actually hurt by adding friction between you and a potential client who just wanted to talk to a human.
Also: if you're not willing to invest the time upfront to train it properly and monitor it on the backend, or pay someone else to do it, don't bother. A poorly trained chatbot is worse than no chatbot at all.
The Training Data Is Everything (Almost)
Please don’t skip over this part or treat it lightly as it’s one of the two things that’s most important in the entire setup of a chatbot.
Your chatbot should be trained only on your content, your FAQs, your product info, your service descriptions, your policies, your brand.
Do not let it wander out onto the internet and pull in general answers, because that's a problem for two main reasons.
First, that information can be wrong. You have no control of what somewhat puts on their website, and it can be inaccurate, incomplete, at odds with what you say, or just plain generic and unhelpful.
Second, and maybe more importantly, it won't sound like you. It won't reflect your expertise, your niche knowledge, or the depth of what you actually know. And once you have someone on your site, you don’t want them wandering off elsewhere.
When we set up the chatbot for our client, we trained it specifically on their product catalog, their FAQ list, their policies, their blogs, and carefully crafted answers to the most common questions their customers ask. That's what makes it useful. That's what makes it feel like them.
The chatbot thus becomes a reflection of your knowledge, and ultimately a reflection of you. So make sure what it knows comes from you and nobody else.
The Prompt Is Just as Important as the Training
The second thing that’s most important is the prompt you give it in the chatbot software.
The training data tells your chatbot what to know. But the prompt tells it how to behave.
That’s an important distinction.
How many times have you encountered a person who’s a “know it all” and tells you that? They just spit out the vast knowledge trying to impress you without really knowing how to say it, and actually how much.
You can also think of it as having a new team member that needs to interact with your clients. You can give them training on all of your products and services in detail, but if you never tell them how to talk to clients, how formal or casual to be, how to handle a frustrated caller, or when to say "I don't know", you're going to get inconsistent results.
The prompt is those directions.
It's a set of instructions you write that guides the chatbot's overall behavior. Things like:
- Speak in a friendly, professional tone
- Keep answers short, no more than two or three sentences
- Always offer a way to contact us if you can't fully answer the question
- Never speculate or guess, if you're not sure, say so and redirect
- Do not pull information from outside the content you've been trained on
Most of the success of a well-functioning chatbot comes from getting both the training and the prompt right
We've seen chatbots that were loaded with great content but gave rambling answers because the prompt was not clear. And we've seen simple chatbots with limited training suggest things way above their capability because the prompt was dialed in.
Get both right, and your chatbot will feel like a natural extension of your business, and thus will reflect well on you.
Set Expectations Right From the Start
This one is simple but critical.
As part of the prompt, your chatbot should tell people upfront that it's a chatbot with limitations.
Not in a scary or complicated way. Just something clear and honest, like:
"Hi! I'm a simple AI assistant. I can answer common questions about our products, services, and policies. For anything more complex, I'll point you to a real person."
Or what we actually used:
“Hello! I'm an AI Chatbot, and I'm here to answer simple questions about COMPANY store. (For complex issues, please follow up by phone or email.)”
This opening message sets the right expectations as people know their questions need to be simple, the answers will be limited in scope, and the answers are from AI, not a human.
In other words, they are not to expect too much and it’s not a human conversation.
So they’ll then act accordingly, and thus build trust.
Keep the Answers Short
Before creating a chatbot for ourselves and clients, I knew we had to address a major shortcoming that frustrates me about the chatbots out there.
I’d ask a simple question and get back a 10-step response with headers and sub-bullets and three paragraphs of background context.
I don’t have time for that! Just give me a couple of things to move me along and if that doesn’t help, give the the capability to contact someone!
So your chatbot should be trained to give short, useful answers. One or two sentences. Maybe a link to more information on your website. That's it.
For example, if someone asks, "Do you ship to the U.S.?" the answer should be something like: "Yes, we ship to all U.S. locations. Contact our team here for a quote: [email]."
Done. Clean. Helpful.
What’s NOT helpful is then asking for the address, weight of the package, size, etc. before giving a response. Because that’s NOT what I asked!
The goal is to give visitors just enough to either solve their problem or take the next step. Not to overwhelm them with everything the chatbot knows.
Plan for the Guardrail Questions
This is something we put a lot of thought into on our client project.
Not every question someone types into a chatbot is going to be a legitimate question. You’ll always have someone that’s frustrating and trying to get answers to questions they shouldn’t get.
For example, people might ask things like:
- "I have a complaint about one of your employees."
- "Can you guarantee this will never break?"
- "Do you have a lifetime warranty?"
- "Can you give me legal advice?"
These are what we call guardrail questions, questions the bot should handle carefully and redirect, rather than attempt to answer directly.
For each of these, you need a specific, pre-written response that's appropriate, professional, and helpful. Usually that means acknowledging the question and pointing them to the right person.
For example: "I'm sorry to hear about your experience. For concerns like this, please reach out to our team directly at [email] so we can help you promptly."
We came up with a list of about 20 of these, and you can use whatever generic AI tool you use to help you create these questions.
And of course make sure you test against them!
Always Give Them a Way Out
No matter how well you train your chatbot, there will be moments when it just can't help someone.
And that's okay.
What's not okay is leaving that person stuck with no path forward.
Every chatbot response should have a clear, easy way for visitors to contact a real person, a phone number, an email address, a contact form link. Something.
We do not recommend adding a live agent handoff feature unless you are a large company with dedicated staff to manage it in real time. For most small and mid-sized businesses, that's simply not realistic. And if you promise a live agent and nobody's there? That's worse than not offering it at all.
Keep it simple. Make it easy to reach you. And let the chatbot handle what it can.
You Need to Monitor It on the Backend
A chatbot is not a "set it and forget it" tool.
You need a way to review what's happening, what questions are being asked, how the bot is responding, and where it's getting confused.
Most good chatbot platforms give you access to chat logs and usage reports. Go through them. And use them for additional training.
This serves two purposes. First, you'll catch problems early, which might be a bad answer, a missing FAQ, a question that keeps coming up that you haven't yet addressed.
Second, you'll learn something valuable about your clients. The questions they ask in that chat window are the real questions they have. That's gold for your content strategy, your sales team, and your product development.
It may take a bit of time at the start to get all of this setup, but it’s worth it in the long run.
Make sure someone on your team gets notified when a chat happens or have someone on a regular basis review the chats. Don't let it run on its own.
Quick Summary: Pros and Cons of Adding a Chatbot
Advantages:
- Available 24/7, thus answers questions even when you're not there
- Reduces repetitive questions to your team
- Captures leads and contact info around the clock
- Gives visitors instant responses
- Reflects your expertise when trained properly
- Can improve conversion if it helps visitors take action faster
- Some people just like to chat quickly versus contact you by email or phone
Disadvantages:
- Requires real upfront work to train correctly
- Poorly built chatbots frustrate visitors, and that frustration lands on your brand
- Can feel impersonal if not set up thoughtfully
- Needs ongoing monitoring and updates, although that decreases over time
- Not right for every business or every type of question
- Information overload in responses is a common and avoidable mistake
So, Should You Add One?
Maybe. But only if you're willing to do it right.
A chatbot that's generic, overwhelming, or frustrating? That's a liability.
A chatbot that's well-trained, honest about what it is, focused on your content, and backed by an easy way to reach a human? That's a real asset.
And that’s what you want to be building in your company. Assets.
We've done this for clients. We can walk you through whether it makes sense for you and your business or organization. No pressure, just clarity. Contact us to know more.
