You never know when you’ll find yourself chatting up a chatbot. I recently went to the SiriusXM support site to find resolution for a streaming issue, when I encountered the company’s support chatbot, Melody. 

Here’s the moment I knew that I was being sent down a fixed set of answer branches with limited generic replies:

A screenshot shows a customer service chatbot interaction

Fair enough, Melody. Only one problem: my issue didn’t fall into any of these categories. So, what was my next step? To query the chatbot with language I hoped would get me in touch with a live agent:

A screenshot shows a customer service chatbot interaction

Eventually, I reached a live chat agent and got my issue resolved. However, it took me more than a couple of steps. And you can see the effort level increasing with each step. Melody, bless its hard-coded heart, is a little flawed.

What if my issue doesn’t fit into the generic categories? What if I’m not savvy enough to know that I can type in “talk to an agent” to get my chat handed off? Wouldn’t it have been so much better if I typed in a natural language query and got connected to exactly the answer I needed the first time, without any more steps required?

Even Salesforce Chatbots Need Help Once in a While

I’m not trying to pick on Melody or SiriusXM. The truth is, a lot of chatbots underwhelm. Usually, the chatbot interaction goes something like this:

A flowchart shows the typical chatbot interaction process
Related readChatbot Technology: Overhyped, or Undersold?

Unfortunately, this is a common experience even for Salesforce’s Einstein-powered chatbots. Typically, the Einstein Bots we encounter are missing a few key ingredients: 

  • No defined purpose and personality
  • Content index limited to only a few sources
  • Limited or no intent detection
  • Lack of personalization
  • Limited search and recommendation capabilities

To improve self-service and case deflection for Einstein Bots, specifically, companies need to do better than this flat “decision tree” experience. The good news is that a little bit of chatbot intelligence can go a long way to achieving just that.

AI Chabot SolutionsPower up your chatbot with AI

How Coveo for Einstein Bots Makes the Intelligent Assist

What’s missing from so many chatbot experiences is contextual relevance. That’s the bread-and-butter of Coveo for Einstein Bots, which weaves enterprise-grade, AI-powered search and recommendations into Salesforce chatbot dialogs.

It’s the best of both worlds, really: when Salesforce Einstein Bot can handle simple issues straightaway, it will; when the customer’s queries “go off script,” Coveo AI can interpret the customer’s language and suggest relevant results.

What’s more, the Coveo for Einstein Bots brings in content from the entire ecosystem—from sources inside and outside of Salesforce. This can include YouTube videos, blog posts, and even community content. Simultaneously, the algorithm considers any other available customer data—such as previous activity, preferences, etc.—to further pinpoint the most relevant content for a given query.

A graphic illustrates how Coveo enhances the Salesforce platform

Keep It Relevant Even When Customers Go Off the Script

Einstein Bot is really effective when customers have a logical task to complete, such as “Login / Password Reset,” “Radio Refresh,” or “Cancel.” For those use cases where the customer has a more complex or nuanced query, Einstein Bot needs the help of a more complex and nuanced relevance engine. A relevance engine pulls together all of your existing content, and searches through it using user intent and interactions.

In my SiriusXM experience, I needed something a little off the script: my streams on the desktop app were delayed and choppy. Here’s how my SiriusXM chatbot adventure might have looked with Coveo’s unified index integrated into the Einstein Bot experience:

A video shows how an intelligent chatbot can retrieve answers from across your content repositories

When I ask a question the chatbot isn’t programmed to answer, it can trigger a Coveo search query and display the top three results. Those search results are based on my query language, profile data, and so on. 

Connect a More Dots Faster, All in Real Time

There was a time when I shared the sentiment that chatbots are dead. It’s all hype! the pundits shouted, rattling their pitchforks. Now that chatbots have evolved to include conversational AI capabilities, I for one have put my pitchfork down.

I’m more optimistic that chatbots can be helpful friends.

That said, I still stumble across pitchfork-worthy chatbot experiences with too much frequency. Fear not, Melody. Keep the faith, Einstein Bots! You’re not headed off to the scrap heap just yet. That is, assuming your creators give you the extra lift you seem to need. 

See, for the Melodies and Einstein Bots of the world, generic playbooks won’t cut it anymore. The time is now to bring in artificial intelligence and machine learning capable of connecting more dots in real time—of delivering up-to-the-millisecond relevance throughout every chatbot experience.

Dig Deeper

Upgrading chatbots is just one of the many ways Coveo can enhance your Salesforce platform. Not only can it bring your enterprise content and data into Salesforce, it can also help employees remain in the flow of work and enhance their proficiency. 

Looking for real-world examples of how the powerful combination of Coveo and Salesforce is transforming service? Check out our ebook, 5 Companies Using Coveo & Salesforce to Transform Service.

Download your copy today5 Companies Using Coveo & Salesforce to Transform Service
Share this story:

About Martin Ceisel

Martin Ceisel is a freelance writer specializing in customer service and B2B. He can be reached on LinkedIn and Twitter.

Read more from this author