I need a fancy new chatbot! Now what?

Chatbots have become an essential tool for many different types of organizations from Universities and Banks to Gaming companies. Typically, the goal is to help better support a customer, student, player or citizen in a more efficient way while also taking some of the pressure off human support teams.

?Generative AI has poured gasoline on this but it has also made things a lot more difficult for buyers to understand.

?The complexity comes in trying to understand what type of chatbot you need and how much you will need to budget for.

?The goal of this article is to share the different types of chatbots and how to assess what you need today and in the future.

?First off, I am assuming you know why you want a chatbot. I am hoping its because you have real problems to solve and not “all the cool kids have one”.

?Typically organizations we speak with want to solve for the following problems:

1. Automate as many requests as possible without hurting the relationship
2. Take pressure off front-line support staff to reduce attrition
3. Encourage more engagement and knowledge sharing with clients
4. Speed up response times and increase availability of support
5. Identify and route quickly to the right human for that personal touch

?There are many other reasons like language support, integration with key systems, analytics, etc. but you get the drift. Now that you know you need one … what type of chatbot do you need and why? Chatbots can be differentiated based on their capabilities, underlying technology, purpose, and the level of interaction they offer. Understanding the different kinds of chatbots and their distinct features can help you determine what you need.

?Before we start though let’s again call out the elephant in the room: generative AI. There is a flood of tools out there including free tools that promise the moon and stars. How can you determine what is marketing fluff and what is real? First you need to get informed.

?How to Determine Your Needs

?When deciding on a chatbot, you need to consider the following factors:

?How much Interaction complexity is needed? If you are looking to support a cascading tree of different experiences for different users, then you will need a more than a generative AI chatbot. If on the other hand you are looking to meet a specific singular need then something like a good old fashioned taskbot might work. If your chatbot needs to assess and help all site visitors including potential clients, returning customers, support requests, partners, potential partners and more you might need a chatbot with a bigger brain and a bigger engine. So, your first step is to determine how complicated the interactions will be and how many different kinds of users you will have.

What integration needs do you have? Do you want the chatbot to seamlessly integrate with other key systems like CRM, KYC, Authentication, SIS, LMS or payments systems (to name a few)? If yes, you will likely want a chatbot that has the ability to do this. Typically this will happen via out-of-the-box connectors and/or APIs or iPaaS providers. It’s important to note that any integrated chatbot will need to be smart enough to know what to do with that data and how it needs to be associated or shared with a variety of user requests. For example if a client wants to know their current balance or the status if a support ticket they will need this. Typically, this would require an NLP or NLU Ai powered chatbot as opposed to generative AI. Generative AI cannot access most secure proprietary systems for security reasons. A connection like this is critical to handle a higher number and complexity of support issues.

What channels do you need to support? Do you need to provide access to your chatbot via SMS, In-App, Web and Social? If your needs are omni-channel it is critical to understand what channels you need to support for your clients. As companies and organizations look to support many different types of clients across different countries and generations they see the need to support many channels. Simpler chatbot solutions typically do not have omni-channel support whereas more complex solutions do support this.

How scalable does the chatbot need to be? Do you need to support hundreds or even thousands of concurrent users? If yes you will need to choose a chatbot that is designed to perform at scale. Typically, lower cost options have limits on access and interaction amounts so avoid these chatbots or taskbots if you want to provide assistance at scale. It is also important to remember that generative AI chatbots typically have a cost for every API call and so every conversation carries a cost. That is why it is also important to understand the commercials behind the solution you choose so that you can manage costs at scale.

What is the ROI you are looking for? ?AI-powered chatbots are typically more expensive than rule-based ones due to their advanced capabilities. These advanced capabilities however can mean the difference between converting a new customer, supporting and keeping an existing customer and reducing stress and lowering attrition on you support team. These benefits can add up into the hundreds of thousands or millions in terms of value. So carefully determine what your goals are for a chatbot when you examine the budget you are willing to spend. Just like anything in this world, you get what you pay for.

What user Experience do you want to promote? Prioritize the customer’s experience, opting for chatbots that offer a balance between technological sophistication and ease of use. Many organizations today want to treat a chatbot as the first line of engagement and support however they also want to identify and route people to agents if appropriate and available. This is critical to balancing the efficiency and access of full automation with the personalization and human touch of live agents. Determine your overall support needs as you approach

Okay so it’s not so easy! I feel its critical to speak with experts that understand how to assess your needs and they can then make recommendations based on those needs.

Now … not to scare you under the couch pillows but there are a bunch of different kinds of chatbots which you can deploy across a variety of situations. Clear as mud? Okay let’s forge ahead.

?

1. Generative AI Chatbots

• Functionality: Generative AI chatbots operate based on large language models (LLMs). The most common is the OpenAI framework (ChatGPT).
• Use Cases: Simple to complex customer service inquiries, FAQs, and guided tasks. Uses natural language prompts.
• Differentiation: These are suitable for businesses with strong websites and documentation.
• Examples: OpenAI (GPT 3 & 4), Google Meena, Rasa, Baidu PLATO, Replika, and Comm100 (using OpenAI)
• Pros: Simple to use, very quick to respond, very fast to train, uses natural language
• Cons: Typically, cannot connect to external systems, provides often bulky answers and can hallucinate or provide incorrect answers.

2. Rule-Based Chatbots

• Functionality: Rule based chatbots operate on predefined pathways and responses. Users choose from given options to navigate the conversation.
• Use Cases: Simple customer service inquiries, FAQs, and guided tasks.
• Differentiation: These are suitable for businesses with straightforward customer interactions and limited variability in user queries.
• Examples: ?Intercom, Drift and Comm100
• Pros: Simple to use, relatively inexpensive
• Cons: Typically, cannot connect to external systems, feel dated, limited in terms of the queries they can handle.

?

3. AI-Powered Chatbots

• Functionality: Use natural language processing (NLP) and machine learning (ML) to understand and respond to user queries with more flexibility.
• Use Cases: Complex customer service, personalized recommendations, and handling a wide range of queries.
• Differentiation: Best for businesses requiring sophisticated interaction capabilities, such as handling nuanced queries or offering personalized experiences.
• Examples: Shopbot by Ebay, Ada health, Mya recruitment bot and Comm100
• Pros: connect to external systems, highly accurate, can leverage libraries of ready intents, feels more human, used for many contexts
• Cons: takes time to train, requires updating

?

4. Voice Bots

• Functionality: Interact with users through voice commands rather than text.
• Use Cases: Customer support via smart speakers, mobile apps, and telephone systems.
• Differentiation: Ideal for services where users prefer hands-free communication or for accessibility reasons.
• Examples: BMWs intelligent personal assistant, LG ThinQ, Samsung Bixby and Comm100
• Pros: provides a different form of interaction, greater accessibility for the visually impaired, good for industries like financial services
• Cons: can struggle with understanding if accents are heavy or there is background noise, slower overall support experience compared to text

?

5. Transactional Chatbots

• Functionality: Designed to assist users in completing specific tasks like bookings, purchases, or form submissions.
• Use Cases: E-commerce transactions, booking appointments, and filling surveys or applications.
• Differentiation: Suitable for businesses looking to streamline particular customer actions or processes.
• Examples: Bank of America “Erica”, Capital Ones “Eno” or H&Ms “Kik” chatbot and Comm100
• Pros: highly focused on and adept at assisting with specific tasks, simple to use
• Cons: limited usage, typically not very adaptable

?

6. Conversational AI Platforms

• Functionality: Advanced AI chatbots capable of continuous learning from interactions to improve responses and user engagement over time.
• Use Cases: In-depth customer service, sales assistance, and interactive engagement across platforms.
• Differentiation: For businesses aiming for high engagement and requiring a bot that adapts and evolves with user interactions.
• Examples: Google Dialogflow, Amazon Lex, IBM Watson and Comm100
• Pros: self-training, data focused, human-like interactions
• Cons: requires training data, requires oversight to ensure accuracy

?

7. Customer Service Bots

• Functionality: Focus primarily on resolving customer issues, answering questions, and providing support.
• Use Cases: Support ticket creation, guidance through troubleshooting, and information retrieval.
• Differentiation: Essential for businesses of all sizes to offer quick and efficient customer support.
• Examples: Salesforce Einstein, Intercom, ZenDesk and Comm100
• Pros: provides

?

8. Sales and Marketing Bots

• Functionality: Engage users with the intent of generating leads, promoting products, and driving sales.
• Use Cases: Lead qualification, product recommendations, and promotional campaigns.
• Differentiation: Useful for businesses looking to automate parts of their sales funnel or enhance their marketing efforts.
• Examples: Qualified, Drift, Intercom and Comm100
• Pros: can be connected to lead enrichment systems, can assist with scoring, can help route incoming leads
• Cons: used only by revenue teams (sales and marketing) as opposed to support

Some chatbot providers (ahem) have chatbots that can do ALL of these things and can map your needs to a cost that is reasonable based on what you want to accomplish.

?If you have read this far … take one additional important step and reach out to me. I am more than happy to listen and provide advice on the subject.

Thanks for reading!

Best, Phil.