How much does a chatbot cost?
A story of sweat and love that goes into building a chatbot and what it might cost.
“So how much does a chatbot cost?”- I get asked this a lot. And the answer is always “it depends…”
This is not something that can be answered like “How much is a pack of milk?” With us, for example, a bot can range anywhere from 2000 USD to 10 000+ USD for setup fee and 100 USD to 5000+ USD monthly retainer, depending on what the chatbot does.
Most Popular Chatbot Tutorials
1. Write a serverless Slack chat bot using AWS
I’m going to share with you some of the most important things that go into understanding how the price of a chatbot is determined and also throw in some example pricing models.
This should give you a good understanding of what to look out for when someone quotes you one or the other price for building a chatbot.
Several great articles have been written already that show how much certain chatbots cost (Great article written by Michael Lucy from AVA here ). I’m going to talk more about what goes into building a bot and setting it’s price. This will help you know what to expect when getting a bot and as well as understand if someone is trying to rip you off or doesn’t know what they’re doing when quoting a price.
Think of it as a guide that tells you what to talk about with a potential chatbot vendor prior to starting cooperation. if you’re developing bots for a client, then this could be good a list of things to go over with them before starting to work together.
As a rule of thumb, the price of the chatbot is usually a function of the work that goes into building it + the value it creates (SaaS/BaaS platforms that provide bots are a bit different in terms of their pricing model. I’m not going to get into that right now since most of the bots built are still very custom)
Here’s an outline of things that i’m going to cover that determine the price of the chatbot:
- Design complexity of the bot
- Chatbot’s AI capabilities
- Technical integrations
- How a chatbot is built
- Who builds the bot/market conditions
- Project management overheads
- Infrastructure costs
- Chatbot launch/Post launch support
By the way, if you’d like to check out other learnings that I’ve published on the topic of chatbots and building a chatbot business, you’re welcome to talk to my chatbot assistant, who will gladly share them with you as well as help you get in touch with us if you need help with getting your own chatbot. You can find her here ;)
Chatbot design and it’s complexity
Not all chatbots are created equal.
To understand what sort of work goes into chatbot design and what makes it complex, let’s take a quick look into how chatbots are designed. Basically, from a design perspective, a chatbot looks like a tree with branches. Each branch represents a conversation flow that needs to be designed. In essence, the branches represent a set of rules like “if chatbot sends this content and the user clicks this button, then this is what happens next”.
As soon as you have more than a couple of conversation branches, you also need to figure out how to connect those branches among themselves so that everything makes sense. The more complex the tree, the higher the price.If you’re just launching your first bot, it’s a good idea to focus on 1–2 main goals that the bot needs to achieve. This keeps the complexity of the bot down and also the price.
Based on what the chatbot is supposed to achieve, the level of complexity can vary from a simple chatbot that is linked to your Facebook page, allowing you to send broadcasts (basically same as email newsletters), to a complicated agent that operates on the same level as a mobile app or website. For example, setting up a newsletter chatbot where the only function is to send broadcasts to users, can be done in less than an hour. All you need to do is set up a welcome message, option to unsubscribe and a few user attributes. You can do it easily using Manychat or Chatfuel. Pretty much anyone can set it up themselves or have a professional set it up for a low fee. Doesn’t require much expertise.
On the other hand, a more complex bot that has several conversational trees, requires good understanding of users and their goals, the business itself, its objectives. This requires great knowledge of conversational design and a certain level of experience and expertise. These factors tend to drive up the price.
If a simpler bot where the only goal is to get users to schedule a call with you can cost roughly 1–2k, then a bot that replaces all the functionalities of your website with sections of “about, products, contact, images, videos, FAQ etc” can start from 3–4k and can go up to 10s of thousands.
Chatbot’s AI capabilities
When we’re talking about AI in conversational bots, then it’s usually in the context of the bot’s ability to process natural human language (NLP for short). This means that the bot is able to understand (to a certain degree) what the user writes to it.
Most of the chatbots that you can see online today, do not have any real AI (Artificial Intelligence) behind them. Usually the user operates the chatbot by clicking on buttons that take them to the next conversational piece, or there’s a set of keywords that the user can use to navigate the bot.
Setting up keyword based bots is relatively easy using platforms like Manychat or Chatfuel.
You just need to define conditions like “if user says “xyz” then the bot answers “zyx”.
This is sufficient in most cases and frankly it wouldn’t make any economic sense for the client to pay for development of true AI. The more keywords the bot has to understand, the more time it takes to develop it, therefore the price increases.
However, if you want the bot to actually understand text and conversations like another normal person would, things get much more complicated. This is especially relevant in customer service bots that need to be able to understand questions like “What are your opening hours next week?”, regardless of whether the user asks it “are you open next week?” or “what time do you open next week”. In this situation it won’t matter, exactly how the words are written in the sentence, but rather the algorithm and advanced language models try to understand what intent the user is expressing when asking his/her question.
In order to teach the AI to understand these kinds of questions, it needs to be trained with a large enough amount of data. It can be rather simple like training the AI to understand and answer only 5 most popular questions that people tend to ask your business, or it can be as complex as Siri that can almost understand and answer any kind of question you throw at it.
Collecting the data and feeding it to the machine learning algorithms + training them adds a significant amount of time to the development process.
You can get a simpler bot that can understand 10 most popular questions people ask for roughly 4–5k USD, built in DialogFlow. From there it can go up to 10s and hundreds of thousands dependent on complexity, additional languages, handoff capabilities, CRM integrations etc…
Btw, you can find out if it would make sense for you to get a true AI assistant for customer service purposes here. We’ve built a little tool that should give you a good indication of how much you could save(or not save) using customer service AI.
Technical integrations/webviews
Another important factor that significantly affects the price of the bot is the amount of technical integrations that are need in order for the bot to fulfill its purpose. For example, you might want to collect emails and phone numbers inside the bot and store them somewhere (spreadsheet or CRM). To do this, you need to integrate the chatbot with another platform via API’s or webhooks.
In most the cases it’s not too complicated and can be done via a Zapier integration in a matter of minutes or hours. This might only add a few hundred bucks to the bot.
On the other end, you might want the bot to pull the data from somewhere else, and for that custom integrations are needed. varies case-by-case and can become a big part of the whole project.
Bots with “standard” integrations (via Zapier or similar tools) can start from like 2000 USD while bots with more “custom” integrations usually don’t start under 4–5k USD.
How a chatbot is built
This ties very much into the first 3 points. A bot can be built in many different ways. For example it can be built from scratch with custom code or using platforms like Chatfuel or Manychat.
It also depends in which platform the bot is going to live on, and the availability of tools that can be used to build bots on that platform.
For example Facebook Messenger is the most popular platform for bots and there are tons of great tools that can be used to build bots for it. If you want to deploy a bot in Slack or Intercom, then the availability of tools is more limited and therefore more work is involved in building one.
Who builds the bot/Market conditions
Like with any other product/service, the market and who is the vendor determines a lot of the costs. There are many vendors who can build a simple lead generation bot, however as soon as you need anything a bit custom, the number of capable players reduces significantly.
The credibility of the vendor is also important, there are lots of people who are jumping into the bot business, providing courses or promising to deliver bots for a small fee. It makes sense to ask them for their past work, read their case studies and see what sort of value they’ve been able to generate for their clients in the past. You don’t want to be paying for someone else’s learning curve.
Project management overheads
Implementing content into the bot usually takes 20–40% of the overall work. The rest goes for technical integrations, user testing, back and forth communication between the client and the bot developer, etc.
Oftentimes you need to revise the content of the bot several times, receive feedback — all those little changes take time and that should also be accounted into the cost of the bot.
When working with larger enterprises, there area lot of things that don’t affect the end user but still that have to be taken into account (privacy policies, data storage, security etc). This also increases the price of the bot.
Infrastructure costs
This one’s pretty straightforward. The infrastructure costs of the bot have to be taken into account as well. Most of the times bot developers use a set of software to put them together, sometimes the software is paid, sometimes a PRO plan is optional. Understanding and knowing what’s involved is important.
For example, if the developer uses Chatfuel or Manychat, then their PRO plans start from 30 USD and 10 USD per month. If you need to forward data to a CRM and need Zapier for that, then that adds another 20 USD per month to the chatbot costs. In case data needs to be stored or moved between servers etc, that has to also be taken into account.
Chatbot launch/Post launch support
It’s important to consider what happens after the bot is built. Once the bot is ready, it needs to be launched. Several things have to be taken into consideration, for example how do you drive traffic to your bot, is it through Facebook ads or a website or some other source?
If it’s facebook ads, the bot needs to be linked with the ads correctly, if it’s a website, it needs to be embedded on the website, etc. This again adds to the work that needs to be done and should be discussed before starting to work together.
Another aspect is what happens after the bot has been launched. It is almost never the case that everything is perfect and nothing needs to be changed. Based on how the users interact with the bot, you see and understand things that need to be changed. Maybe you even want the bot developer to send out user broadcasts for you? Or does the developer give you a crash course on how to do that yourself?
If it’s an NLP bot for customer support, then we always recommend to monitor it for at least 3 months. During this time improvements are made, the bot gets smarter and more problems are solved.
Ongoing support for conversation analyzis, bot training, monitoring and fixing is usually done on a monthly fee.
Recap
So there you go. These are the things that I think must be considered when thinking of getting a chatbot. It might seem like a lot to take in, but in reality it’s much simpler when you’re already discussing things with your potential bot developer. They should be able to walk you through the process and give you clear guidance.
I hope this has helped you understand more what sort of work goes into developing and launching a bot. If you’d like to see other case studies and posts on the topic of chatbots then our bot is happy to share them with you. For example, we recently published an awesome case study that shows how a real estate lead-generation bot helped sell 3 apartments in 10 days. Just hit THIS link to check it out ;)
Guide created by Juhan Kaarma, co-founder of ChatCreate. We build chatbot solutions ranging from simple lead generation bots to complicated Natural Language Processing bots with custom integrations. Companies we work together include small businesses, government organisations as well as enterprises.
Feel free to reach out to me on Linkedin or juhan@chatcreate.com if you need help building a chatbot or need consulting on building a chatbot service.
Liked that guide?
Smash and hold that “👏” button and watch those numbers roll. If you’d like to be the first to know when we publish similar case studies, subscribe here 👈
