While the most basic of chatbots may be perfect for your needs, this article is going to cover the fullest extent of what chatbots are capable of, how much they cost and what to expect in the future.
Hopefully, if you decide to adopt or upgrade your automation program within your business, you’ll have more realistic expectations and won’t go overboard with elaborate AI capabilities.
So, how many different types of chatbot are there?
You can boil chatbots down to 5 different types:
- Text only
- Voice only
- Chatbots that learn
- Chatbots that don’t learn
In our humble opinion, we find a more constructive way of looking at this would be through maturity levels of a chatbot. By understanding what is powering your chatbot, we can group together the types of chatbot by capability and overall usefulness to you.
- How chatbots have evolved over time
- The chatbot maturity model
- What chatbot is right for me?
Let’s start off with the basics and understand the evolution of chatbots:
How chatbots have evolved over time
1964 – ELIZA
ELIZA was born at the MIT Artificial Intelligence Laboratory by Joseph Weizenbaum, Eliza gave the illusion of understanding and was the first program to attempt the Turing test.
2010 – Chatbots Reemerge
Thanks to advancements in computing power, software developments and APIs, basic chatbots started to remerge, responding to simple FAQs to reduce repetition and increase productivity.
Fed with a static and clearly structured database. Any complex issues were immediately handed over to human.
2016 – Chatbot Hype cycle begins
It’s at this point marketing got their hands of this technology and decided it would be the next big thing. “The end of apps” was mentioned more than we care to remember.
While chatbots were over promised; capability did increase with the ability to:
- Broadcasts messages
- The emergence of omnichannel & hybrid interfaces
- The emergence of NLP tools with business-user friendly front end interfaces (Such as Dialogflow and Chatfuel, which are in part to blame for the hype cycle as building chatbots became accessible to non-developers)
2018 – Digital Assistants
Assistants are able to have contextual conversations, carry out tasks such as proactively schedule meetings and booking cabs.
2020 – Bot-to-bot
For 2020, we predict that chatbots will start communicating and sending commands to other chatbots and working increasingly closer to RPA tools.
The Chatbot Maturity Model
Level 1: basic
It’s these types of chatbots that risen from the chatbot hype in 2015 and are also the reason for chatbots poor reputation that they’re annoying assistants with limited capability.
These basic assistants are great when you keep the focus on what you want to automate narrow and clearly defined. They are great at handling short head questions, which make up 70-80% of common enquiries.
These chatbots will either reply to very simple FAQs that only have a few ways of asking that question.
You can build conversations flows, but they will need to heavily rely on buttons and cards.
Most of the time, a human agent will need to be ready and available in the likely even the bot doesn’t understand a user input.
Example conversation flow
Popular use cases
- Automation of simple & basic FAQs
- Personal reminders
- Links to further information
For more examples and use cases, it’s well worth visiting our comprehensive list here.
Typical tools used:
Cost range: free – £10,000
Good news! Cost start from completely free. Although the likelihood is you’ll need to pay a monthly subscription. Some platforms can be hard-hitting but are well worth the investment.
If you’re looking to use these basic chatbots for things like lead generation as part of your marketing funnel, then we’d recommend bringing in a consultant as it would save you a lot of time and pain along the way.
Level 2: semi-scripted
The jump between these two levels of chatbot is pretty dramatic. Not just in terms of capability but cost as well. This is the phase that follows on where most businesses are currently at: the proof of concept stage.
The complexity grows for a multitude of reasons. To grow the capability of a chatbot means you need to start integrating with other back end systems and collecting more user information. Secondly, you will want to start handling more complex sentences which require natural language processing tools.
We go into more detail about the inner working of a chatbot here.
Example conversation flow
Popular use cases
- Banking transactions
- E-commerce product recommendations
- Healthcare support & diagnostics
Typical tools used
Investment range: £10,000 – £300,000+
Sadly, the skillsets needed to successfully create and manage chatbots on tools like Dialogflow are still hard to come by. Thus the cost bumps up significantly.
Secondly, it’s imperative thanks to the likes of GDPR that handling user data and integrations into other backend systems is done correctly.
The investment is obviously still worth it, as we near 80% of companies adopting and progressing their chatbot capabilities.
We go into the costs of building a chatbot in more detail here.
Level 3: contextual
As you will have experienced first hand – typing into a chat window freely has a low chance of having the chatbot understanding every request. Especially if you ask something unexpected mid-flow.
The difficulty in part lies in how we humans communicate.
Often we talk with a background of context.
For example: If we’re friends and you say to me: when is that concert on? There are multiple underlying bits of information that need to be understood before we can answer that question correctly:
- Which concert do you refer to?
- What location?
- What timezone?
and so forth.
Rasa has a great example of how much ambiguity and information can be contained in a simple conversation.
Check out their in-depth explanation: what is a contextual chatbot?
Example conversation flow:
Level 4: Fluid assistants
Level 4 fluid chatbots are currently the most cutting edge of chatbot technology. This standard is only achieved by the dedicated chatbot and AI companies trying to solve a few specific problems.
Some great examples are the likes of Babylon and Ada. These chatbots are multi-modal (meaning they can be used on multiple channels and interfaces.
Secondly, they are also implemented with an app-style interface to improve the user experience even further.
Most enterprises are still 2-5 years away from even matching the hybrid design and experience of dedicated apps such as Ada.
The final piece of the puzzle before chatbots can progress to the next level would be seamless integrations with other apps.
Imagine for a moment, that when you asked Siri to put a task in your Google Calendar or buy a ticket to an event on Eventbrite.
While technically viable, we have a long way to go to for now. As we mentioned since, it’s only companies with a very specific use case, these bots are a result of companies going through multiple rounds of investment.
To give some context, Ada, to date, has raised $69m over 3 rounds and ploughed that all into one user experience.
Example conversation flow:
Level 5: Automation hives
It’s here we enter the fun part: a world of hypothetical and wild speculation.
How technology evolves over the next 5-10 years is anyone’s guess.
It was only a year ago we believe very creative practices such as writing articles was a thing humans will only be able to do for the foreseeable future.
And then GPT-2 came along and blew that out of the water.
What we are certain of, is that in the very-near-future, the lines of robotic process automation (RPA) and chatbots will blur.
The only thing left for chatbots to do once they become integrated into most of your apps and software with flawless conversation design, is to start working with each other!
We advise all the business leaders we consult that AI assistants will lead to a significant shift for society, with many implications for businesses and their customers.
Leaders who can see beyond short term from and goals will benefit from this capability and will be crucial for survivability.
Additionally, we predict there will be thousands of layers of automation with machine learning algorithms designed to manage and orchestrate all these individual systems.
The human will become the decision-maker of a robot manager. Manging robots that talk to each other.
Exciting times ahead.
Level 6: Artificial General Intelligence (AGI)
Also known as the technological singularity.
I appreciate for some, you may consider this quite a jump from “automation hive” to “AGI”. Our maturity model is something we will continue to refine and improve as soon goes on the future becomes less hypothetical!
So we’ll gloss over this level quickly as it also becomes a heated debate of ethics and philosophy. Coated in the fact there’s no guarantee we’ll ever achieve this at all.
We have enormous hurdles ahead such as increasing computing power, handling ever-growing datasets to make marginal gains of accuracy.
Many debates the current techniques we use for deep learning are already reaching the upper end of their capability.
What type of chatbot is right for me?
Picking the right chatbot for you depends on a huge amount of variables such as your budget, how much data you have and the size of the company.
Or feel free to ask our community – they’re a friendly bunch that will happily help find the right answer.