E-commerce platforms and streaming services have made us citizens of digital ecosystems – spoilt for choice with endless selections and pampered by digital assistants. With instant comparison options, you can base your purchasing decisions on competitive products, features, prices, discounts, and customer reviews.
You’d be tempted to call these interactions as “searches” – but they are so much more. These data-driven, proactive suggestions are provided by “recommenders” that bring you personalized experiences – based on your previous searches, purchases, and clicks – rolled out as a digital red carpet where you are shown options based on your proven (or potential) interests.
The Evolution of the Personal Shopper in Digital Services
When you look for recommenders in the brick-and-mortar world, the closest analogy I can think of is the personal shopper in a luxury retail store. This store employee is essentially a salesperson who knows their clients’ preferences, styles, and budgets.
These “human recommenders,” however, have their limitations. While they are good at providing highly personalized shopping experiences, they cannot, match the scale or wide range of parameters offered by digital recommenders. Your Netflix profile is a great example of the power of digital recommenders: you are prompted to check out or bookmark films of the same genre, same cast, same franchise, similar plots, or movies trending in your country. You have the option to post your likes or dislikes – which in the background, trains the recommender on your unique preferences.
Support Recommenders in Brick-and-Mortar Experiences
Recommenders are not limited to the purchasing experience – they extend across the customer lifecycle right up to post-purchase support scenarios. Rewinding to the eighties and early nineties, a support scenario for a damaged product would typically comprise a series of interactions with “human recommenders” – who are representatives from different departments of a physical retail store.
Imagine that your used bicycle suddenly develops a defect in the brakes, and you take it back to the seller. The employees at the reception would address you by your name as you have been a loyal customer for years. They are aware that you have purchased premium products before, and you are immediately offered a repair option. While it is clear that the malfunction is from wear and tear, the support employee continues to offer you a free repair – in view of your purchase history.
Support Recommenders in Digital Experiences
A significant part of digital purchases involves adoption, replacements, or returns. Which brings us to the support experience extended to online consumers.
Digital recommenders are driven by data captured across multiple customer touch points. The recommendation systems provide personalized, relevant, and quick suggestions for resolving customer issues.
Examples of customer support recommenders include:
Knowledge Bases: The system proposes FAQs, knowledge base articles (KBAs), and self-help resources either through natural language processing or keywords.
AI-Driven Chatbots: AI-driven, automated conversational agents provide responses and solutions based on learned data.
Past Case Resolutions: The system proposes the best resolution based on patterns observed in past cases and issues faced by similar customers. The system also recommends workflows for support engineers, to improve case resolution.
Escalation Paths: The system offers escalation paths based on sentiment analysis and severity of issues, ensuring quick resolutions.
Ticket Categorization: The system classifies support tickets and automatically assigns priority levels and product categories based on the severity or impact of the issue.
Example: AI-Infused Support Recommenders in Business Software
As a leader in enterprise software, SAP caters to a wide range of markets, industries, lines of business, and customer segments. The scope of providing support, therefore, extends to a heterogenous customer base that spans multiple subscription tiers, sizes, and business impact categories.
For example, let’s take a look at SAP’s Get Support application, which uses Incident Solution Matching (ISM), an AI-integrated service based on machine learning and large language models (LLM) – to improve its real-time support capabilities that address a heterogenous customer base.
SAP’s Incident Solution Matching is a knowledge base recommender, where the service automatically proposes relevant solutions from SAP Notes, SAP KBAs and SAP Help by analyzing case data provided in the Get Support application.
When you pop open the hood, you find that ISM uses proprietary AI models that apply a semantic retriever technology to boost the relevance of recommendations. The semantic retriever, coupled with a semantic ranker, helps the system prioritize solutions most likely to resolve a customer’s issue.
With the help of LLM-infused models, ISM now delivers an improvement of over 25% in recommendation relevance and accuracy – ensuring faster and more personalized solutions for customers. The performance of these models is improved on an ongoing basis: from the first version to its current 11th iteration, SAP models have improved by 2.7 times.
What does this mean for SAP customers? SAP’s Incident Solution Matching is a significant step towards improving the support experience for customers. The enhanced recommendation system allows SAP users to resolve their issues more efficiently by reducing downtimes and times to resolution, while increasing productivity. These enhancements significantly improve the quality of personalization, self-service, and agent productivity.
In the end, all of this translates to a high-quality, seamless support experience that can be accessed directly from the customer’s support account.
