Integrating AI-powered chatbots and search on your website

By Janus Boye

Nicole Rogers is co-founder at "Generative AI Copilots" - startup ai12z

Across all industries, organizations are exploring how to leverage the power of Generative AI. Everyone is talking about AI today, but how can you take advantage of this in the content management world? How can you leverage all the content you’ve created and combine that with AI to provide a great experience for your customers?

In a recent members’ call we heard from Nicole Rogers from "Generative AI Copilots" - startup ai12z. Nicole shared how Generative AI can be combined with your content and your CMS to enhance your site visitors’ web experiences. Examples of how you can incorporate this technology right on your website includes:

  • A chatbot in the corner of your website to answer your site visitors’ questions about your products and services.

  • Digital assistants (aka “copilots”) to help them along their journey, providing personalized recommendations and assisting in completing tasks like making a purchase.

  • AI-powered search to respond to your users’ search queries with detailed summaries.

Nicole is co-founder at ai12z and previously worked in senior marketing roles at WeWork, Google, Box and Ektron. ai12z has a partner model and works closely with digital agencies to enable customers to quickly launch copilots.

Our conversation started with looking at copilots and how it’s the next big thing.

Every company will have many copilots

As we’ve heard in previous conversations, Copilots is not limited to Microsoft and is becoming a standard industry term. Nicole used this simple definition:

An application that uses AI and LLM to help you perform a complex task. It has a conversational interface.

In other words: An AI-powered chatbot.

Nicole shared how Copilots can be used for websites across industries to help support prospects, customers, students, citizens and whoever else is using your website.

Imagine ordering a blue sweater, but getting a green one instead. With a copilot on the website, you don’t have to navigate the site, identify returns on a crowded site and read through to find the right link. Using a simple conversational interface, you could easily get a prepaid return label and the blue sweater is already on its way with express shipping!

Similarly, copilots can also be used for internal teams. Nicole shared examples from customer support, sales enablement, HR and IT. Consider common internal tasks and questions like:

  • Can you help me prepare for my sales call?

  • Can you sort through the resumes for the product manager role?

  • The app is running slow. What can I do to speed it up?

So, how does this really work?

Copilots under the hood

Using the below illustration, Nicole shared an outline of how the technology works. Notice how you can rely on established large language models or even bring your own.

When a user asks a question, the question gets turned into content embeddings or vectors, which are then compared to the most relevant content in the organisation's database. The content is then reranked, and the top three to four most relevant documents are used to create a coherent answer using a large language model. An advanced copilot goes beyond just a regular chatbot system and automates workflows to provide a better user experience.

For instance, in a retail example, a user wanting to return shoes would have the copilot call a CRM connector to check when that user bought the shoes. The copilot would then check the return policy and execute a plan to process the return. The context engine and AI conversational history are also crucial, allowing the chatbot to understand the user's previous questions and provide more accurate answers. With purpose-specific copilots you can get even better results.

She also introduced us to the use of retrieval-augmented generation (RAG), one of the hottest buzzwords at the moment. RAG allows the system to access and understand an organisation's content to provide more accurate and relevant responses to users.

What ai12z is offering is beyond a RAG system, as it’s not just answering questions, but also capable of automating workflows and using tools (plugins) to further extend the functionality.

Using copilots to make search better

An example of how TSA Pre search could work with new search. Click for larger version

AI-powered search is different than traditional search as we know it. As Nicole said, while some websites have effective search functions, many require users to sift through multiple results to find the answer to their question.

Nicole showed how AI-powered search can streamline this process by filtering results and providing an answer directly. Users will be able to rely on semantic search rather than just keywords. For instance, you can ask the search bar a question (instead of keywords) and the AI-powered search will understand what you’re asking. It will then present an answer, in addition to the long list of search results we’re used to seeing today.

Imagine the following on TSA’s website, where you can ask about TSA Pre. You can even ask a follow-up question and it will understand that “it” refers to “TSA Pre.”

Users can even ask a question in their language of choice and the AI-powered search / copilot will understand what the user is asking and answer in their language.

An important point is that as an organisation, you can control how the AI responds and its ability to only answer questions related to that organisation’s specific content. In other words, the copilot will not hallucinate or make up things. You wouldn’t want your copilot to be talking about your competitors. So you can actually set guardrails or boundaries so that it only answers questions based on your organisation’s content.

As a positive side effect: This can actually also be used improve your customer experience and also SEO, as you can identify which questions have not been answered and by addressing unanswered questions, you can discover what content your visitors are seeking (which is not currently on your website). Creating that content will serve your audiences better and also ultimately boost your SEO.

Learn more about AI, chatbots and better search

The conversation about innovative solutions and real-life implementations of Copilots naturally continues in our peer groups and conferences. You can meet Nicole in person at our upcoming CMS Expert sessions and expect to see this topic on the agenda in the foreseeable future.

You can also join us in Montreal in August for CMS Connect 24 or next year in January in Florida for CMS Kickoff 25.

Our Canadian member Seb Barre at TELUS has generously hosted two member's’ calls on his learning progress:

And we’ve also heard from Tom Cranstoun (aka “The AEM Guy”) on Rethinking AI’s Role in Content Management.

Finally, you can also download the slides (PDF) or even lean back and enjoy the entire recording from the call.