It's 2024, which means almost every SaaS product has a chatbot or conversational AI tool in the bottom right-hand corner.
Everyone knows that user assistance tools can be instrumental in helping reduce support tickets, answer user questions quickly, or at least more accurately route a user to the right support agent. However, not all user assistance tools are the same.
There's a huge difference between the chatbots of yore and today's conversational AI tools.
It's important to understand the differences between these products and determine which is best suited for your user's needs.
History of chatbots and conversational AI agents
Let's quickly review both conversational AI agents and chatbots.
Chatbots have been around for decades. They're much simpler and have traditionally relied on decision trees, where one response leads to a few more options for response and then to another choice of options, which leads to a few more options and so on until the solution, or an agent, is reached.
This actually was exceptionally helpful and revolutionary when it came out because it helped automate things like call routing on the phone or assisting users through a chat-based tool. Think of Facebook's chatbots in the 2010s, which helped businesses quickly route and address customers' needs. Instead of having to actively respond to the message, the chatbot could use its logic tree to ask users to categorize their problems and then funnel them into the right flow. The issue is, at the end of the day, these interactions are fairly static and scripted and don't have the dynamism needed to really address users' needs, especially in the more complex world of B2B SaaS.
Conversational AI
That's why the conversational AI agents that have appeared over the last two years since the popular release of several powerful LLMs has been a huge development not just for customer support, but for user engagement in general.
These conversational agents use their massive amounts of data and machine learning to power their decision-making and understand the user's incoming questions and respond with a contextual answer.
AI support agents in particular, like our Copilot, have also been trained to take on a variety of voices and personas that made this natural language understanding feel even more human. Most folks' earliest interactions with a tool like this were with Amazon's Alexa platform or Siri on your iPhone. And as great as these were when they were first released, they now are fairly clunky and dated compared to what folks can do with tools like ChatGPT.
Another advantage of conversational AI tools is that they can actually learn as they go. Unlike a static chatbot, as you talk with the conversation AI tool, it's able to learn about your problems and fix them, take in your feedback and store it in memory, and use it for the future.
How to choose
So, it's clear that there are big differences between conversational AI agents and chatbots. But what is the best fit for your product? Let's run through a couple of examples.
It's easy to rule out chatbots completely and decide that you're going to go for the best conversation AI agent. But there is a reality that there are some workflows in which having a simple chatbot might actually be easier than having a highly smart and trained conversational AI agent.
For example, it might be helpful to have a simple chatbot which can handle your most repetitive and clearly order tasks.
Or, you might use an initial chatbot interaction to guide users into a particular funnel before handing them over to an agent. The less complex your tool, the more likely this can be a good solution.
But why would you ever choose something like this?
Well, it's cheap.
In fact, even as costs continue to go down for the powerful AI agents, at scale they can still be quite expensive.
So if you have a fairly non-complex tool with 80%+ plus percent of your incoming queries being regular and predictable, you can actually rely on a chatbot to handle this and save yourself money.
The reality
However, I don't think that's the case for most B2B SaaS tools, particularly those serving enterprise level. The reality is that in 2024 you should probably be leaning towards a fairly powerful conversation agent which has all the advantages of the large language model behind it.
Because this agent can understand your users' questions in context, and sort through all of its knowledge as well as your training of the model continuously, it can answer, get feedback, and learn.
It's like a support agent who is on 24/7 and is super invested in improving their skills and knowledge.
How to best deploy a conversational AI tool or chatbot?
So how do we actually deploy a conversational AI tool the right way?
Well, we wrote a full post about this, which you can see here. But the crux of it is that you need to make sure that you understand clearly the use case for this conversational agent before you go out and get it.
A coherent strategy and vision for how it will incorporate into your customer success team is critical. You can't just assume you can train it on all your models on all your knowledge base I should say and then fire your entire support team and let it run.
At this stage, it's still a complimentary factor which is best suited for quick answer or routing to agents.
But there are marked advantages of a conversation AI which can't be denied.
They're on 24/7 and even though they might still be expensive at scale, they are minuscule compared to the team of agents that you might have needed to have previously.
Don’t forget this
One of the biggest potential hurdles that folks don't seem to recognize when they're considering implementing a conversational AI tool is how important it is to have great source documentation. What do I mean by this?
Well, imagine that tomorrow you go out and buy a great AI assistant.
You're excited to launch it into your tool, so you onboard it and connect it to all of your help documentation and then get it live.
Well, on day one you get a lot of engagement with it which seems great, but as the days go by you realize that a lot of the queries are being answered with outdated information, incorrect answers, and you're usually getting frustrated because it doesn't match up with the actual product experience at this time.
What's going on here? If your documentation is overlapping or confusing or outdated, the ansers are going to be off, and that's not of much use to users — bad UX!
When you go to implement a conversational AI tool for support use case, or really for any use case, you want to ensure that the data and knowledge that you feed into it is 100% accurate, timely, and most importantly, that it will be updated as your product progresses. You need to have someone owning that responsibility to update your documentation particularly if your company launches fast like so many startups.
What we’ve seen
It's pretty obvious that a conversational AI agent is extremely robust when compared to a chatbot. But as we discussed, that doesn't necessarily mean it's the best fit for your business, particularly if you're on a budget or if you have a high volume percentage of questions that are fairly repeatable and static. In that case, you could lean towards the chatbot.
One thing we've seen on the ground here at CommandBar as we've worked with clients to deploy our conversational AI support agent, Copilot, into their product is that the quality of response is fairly high off the bat for folks who have made the proactive effort to ensure that the data they feed into it is up to date in their documentation.
When this is done, we've actually found the model doesn't need that much training, usually less than a week or two, which is huge. Folks have gotten up and running really quickly and launched to users with confidence.
Once launched, they've seen increased user engagement with Copilot, as well as reduced tickets and more overall user satisfaction. It's pretty exciting to see this, but we know that this is not the end. We continue to update Copilot and work towards creating a best-in-class user assistant that can serve both customer support and on-app messaging function.
While both conversational AI agents and static chatbots have potential use cases for B2B SaaS businesses, I would generally recommend that folks go with an AI agent as long as they have a well-thought-out strategy for deployment and upkeep of their source documentation. The AI agent simply has so much more power at a marginally higher cost, and will also be able to drive user insights for you, like query analysis or fallback funnel analysis, which can identify where your source documentation is weak and where you need to add further context.
