Get to Know Einstein Article Recommendations
After completing this unit, you’ll be able to:
- Describe Einstein Article Recommendations.
- Explain the key benefits of AI-powered customer service.
- Prepare your data to make accurate predictions.
Meet Einstein Article Recommendations
At Ursa Major Solar, a Southwest-based supplier of solar components and systems, innovation is a large part of their success in making the world a greener place. The company recently released a new line of products, ranging from a supersonic solar panel that energizes cars to individualized solar energy converters for laptops. Many of the company’s customers are rushing to try out these new products—but those that aren’t as tech-savvy seem to be a little intimidated. To get all of her customers on board, Ursa Major Solar’s CEO, Sita Nagappan-Alvarez, knows that her support team must be ready with info and guidance to make customers feel comfortable.
Luckily, Ursa Major Solar has a bunch of knowledge articles about their products and services. But Sita isn’t sure how to help her service team quickly find the right articles to solve each customer case. Maria Jimenez, her service admin, has a solution: Einstein Article Recommendations.
Einstein Article Recommendations uses artificial intelligence (AI) to recommend relevant knowledge articles on customer cases so agents can address customer inquiries more efficiently. Einstein can search hundreds of articles—even hundreds of thousands of articles—in a matter of seconds, saving your agents time and energy. Agents can then review Einstein’s list and attach an article to the case, editing it if needed, or they can dismiss the recommendations and view more.
According to Salesforce research, the benefits of AI-powered customer service are significant. Data suggest that you can increase agent morale by 75% and first contact resolution by 82%. Implement Einstein Article Recommendations, and watch your agent and customer satisfaction grow.
Sita is still learning about AI and has lots of questions for Maria, like: How do AI recommendations work? What determines which recommendations are shown? Do recommendations change over time? Maria is happy to help. She explains that Einstein is a form
of AI that runs on algorithms made for the very specific purpose of organizing their business.
Einstein Article Recommendations, specifically, uses AI to learn and become more powerful over time. This Einstein feature is trained using a form of AI called natural language processing (NLP), a smart technology that can read articles and closed cases to learn how to match them together. The tool’s algorithm also ranks articles based on their relevance to an open case, so Einstein only recommends highly relevant articles.
So, how does Einstein determine article relevance? The friendly AI assistant considers a few factors.
- Team overlap: How closely the case’s language matches the article’s language
- Attachments: How often agents attach an article to similar cases
- Dismissals: How often agents dismissed the article as not helpful on similar cases
- Term span: The distance between certain case terms in the article
- Longest common sequence: The length of the longest common text sequence between the case and article
Sita sees the value in the Einstein features that Maria described, so she gives her the go-ahead to explore further.
What Data Do You Need?
Maria knows that Einstein must analyze a certain amount of data in order to make accurate recommendations. She takes a closer look at the exact requirements.
|What You Need
||Required or Recommended?
|100 or more English Knowledge Articles
||Your article coverage must be extensive enough to address common customer questions. Articles don’t need any special tags, data categories, or metadata to be recommended. Newly published or overwritten articles automatically undergo term analysis
to be considered for recommendation.
|1,000 or more closed English cases
||Einstein looks at past cases to learn how your organization works. These cases need at least one text field that describes the customer’s issue—for example, Description. They don’t need to have articles attached to them, but if they do, that’s even
|500 or more English case-article attachments
||Einstein learns from each instance of an article being attached to a closed case (also called an “attach”). The more attaches you have, the better. However, even with fewer than 500 attaches, your model can still make accurate predictions based
on the content and overlap between cases and articles.
To Maria’s delight, Ursa Major has more than 100 knowledge articles in its Salesforce Knowledge base. She does some more digging to find that her awesome support team closed a whopping 1,700 cases in the past year—and many of them have knowledge articles attached! Maria and her team have met all the prerequisites to get started.
With high-quality data to learn from, Einstein is sure to make accurate predictions and help Ursa Major’s support team—and their customers—succeed.